A contribuição dos programas de pós-graduação multi e interdisciplinares para a produção de tecnologias no Brasil: uma análise exploratória e preditiva
Métricas e modelo: Vivian Alves (tese UFSC), apoio Viviane Schneider. Análise estatística e Código fonte: Viviane Schneider
inicio em 24/7/2020. última atualização em: Versão 4. Data:04/8/2020.
library(dplyr)
library(plotly)
library(ggplot2)
library(tidyr)
library(magrittr)
library(plotrix)
library(rgl)
library(car)
library(lubridate)
library(ggplot2)
library(GGally)
library(corrplot)
library(corrgram)
library(ppcor)library(readxl)
IMI_e_IPT <- read_excel("IMI e IPT.xlsx",
col_types = c("numeric", "text", "text",
"text", "text", "text", "numeric",
"text", "numeric", "text", "text",
"text", "text", "text", "numeric",
"numeric", "numeric", "numeric",
"numeric", "numeric", "numeric",
"numeric", "numeric", "numeric"))## New names:
## * `` -> ...1## [1] 2858 24outliers <- boxplot(IMI_e_IPT$IPT, plot=FALSE)$out
no_outliers <- IMI_e_IPT
no_outliers <-no_outliers[which(no_outliers$IPT %in% outliers),]
graf <- IMI_e_IPT[, 15:24]
dim(graf)## [1] 2858 100.1 Coerência do modelo de análise
1 Testes de normalidade
Com seu papel no teorema central do limite, a distribuição normal é encontrada em muitos dos testes estatísticos chamados gaussianos ou assintoticamente gaussianos. O pressuposto de normalidade é feito sobre uma distribuição a priori em um teste de aderência para indicar que esta distribuição segue aproximadamente uma distribuição normal. Existem vários testes de normalidade.
panel.hist <- function(x, ...)
{
usr <- par("usr"); on.exit(par(usr))
par(usr = c(usr[1:2], 0, 1.5) )
h <- hist(x, plot = FALSE)
breaks <- h$breaks; nB <- length(breaks)
y <- h$counts; y <- y/max(y)
rect(breaks[-nB], 0, breaks[-1], y, col = "cyan", ...)
}
# 1.2 by Melina de Souza Leite
panel.lm <- function (x, y, col = par("col"), bg = NA, pch = par("pch"),
cex = 1, col.line="red") {
points(x, y, pch = pch, col = col, bg = bg, cex = cex)
ok <- is.finite(x) & is.finite(y)
if (any(ok)) {
abline(lm(y[ok]~x[ok]), col = col.line)
}
}
# 1.3 help(pairs) by Melina de Souza Leite
panel.cor <- function(x, y, digits = 2, prefix = "", cex.cor, ...)
{
usr <- par("usr"); on.exit(par(usr))
par(usr = c(0, 1, 0, 1))
r <- abs(cor(x, y))
txt <- format(c(r, 0.123456789), digits = digits)[1]
txt <- paste0(prefix, txt)
if(missing(cex.cor)) cex.cor <- 0.8/strwidth(txt)
text(0.5, 0.5, txt, cex = cex.cor * r)
}
pairs(graf,
diag.panel = panel.hist,
upper.panel = panel.cor,
lower.panel = panel.lm,
main="Correlação Multivariável")
1.1 Dados não normalizados
## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 0.00000 0.00000 0.04546 0.21083 0.21212 5.15000## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 0.2304 0.4178 0.5025 0.5051 0.5868 1.0234## [1] 2858 24
1.1.1 Shapiro-Wilk test
O método de Shapiro-Wilk é amplamente recomendado para teste de normalidade e fornece melhor potência que o K-S. É baseado na correlação entre os dados e as pontuações normais correspondentes.
A partir da saída, o valor p> 0,05 implica que a distribuição dos dados não é significativamente diferente da distribuição normal. Em outras palavras, podemos assumir a normalidade.
Neste caso o teste mostra que a distribuição não é normal nos dados não normalizados.
##
## Shapiro-Wilk normality test
##
## data: IMI_e_IPT$IPT
## W = 0.52019, p-value < 2.2e-16##
## Shapiro-Wilk normality test
##
## data: IMI_e_IPT$IMI
## W = 0.99093, p-value = 1.775e-12##
## Shapiro-Wilk normality test
##
## data: IMI_e_IPT$IPT
## W = 0.52019, p-value < 2.2e-16##
## Shapiro-Wilk normality test
##
## data: IMI_e_IPT$IMI
## W = 0.99093, p-value = 1.775e-121.2 Dados normalizados
norm_MI_PT <- IMI_e_IPT %>%
mutate_all(replace_na, 0) %>%
mutate(IPT_t = 0.510 + log(IPT)) %>%
mutate(IMI_t = 10 + IMI) %>%
filter(IPT_t != "-Inf")
summary(norm_MI_PT$IPT_t)## Min. 1st Qu. Median Mean 3rd Qu. Max.
## -3.8207 -1.9749 -1.0994 -1.0274 -0.1831 2.1490## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 10.23 10.42 10.50 10.50 10.58 10.94## [1] 1496 26
1.3 Verificação de ajuste do modelo
As estimativas são os coeficientes das variáveis independentes do modelo linear (todas as porcentagens) e refletem uma alteração estimada na variável IPT dependente (a qual foi avaliada em sua composição separadamente em Patentes, Produtos e Aplicativos), quando a variável independente correspondente é alterada.
Portanto, para cada aumento de 1% em porcentagem “Produção de Tecnologias (IPT)”, espera-se um aumento ou diminuição significativo (nos resultados em aparecem asteriscos) das variáveis independentes (Y), mantendo todas as outras variáveis constantes.
Variáveis dependente (Y)
Y = somatório(Y1, Y2, Y3)
Y1 = Quantitativo de produção de patentes por Programa (SPPP/QPPP)
Y2 = Quantitativo de produção de produtos por Programa (SPPPr/QPP)
Y3 = Quantitativo de produção de Aplicativos por Programa (SAPP/QPPP)
Variáveis independente (X)
X = somatório(FCDo, FCDi, CC, CP)
X1 = Formação do corpo docentes dos PPG (FCDo)
X2 = Formação do corpo discente dos PPG (FCDi)
X3 = Colaboração Científica (CC)
X4 = Contexto profissional (CP)
X5 = DEPENDENCIA ADM
X6 = CONCEITO
X7 = UFPROGRAMA
1.4 Produção de Tecnologia (IPT= Patentes+Produtos+Aplicativos)
1.4.1 Com outliers
1.4.1.1 Regressão linear
fitIPT<- lm(formula = sqrt(IPT) ~ FCDo + FCDi + CC + CP + as.factor(DEPENDENCIAADM) + as.factor(CONCEITO) + as.factor(UFPROGRAMA) + FCDi * CC - 1, data = IMI_e_IPT)
summary(fitIPT)##
## Call:
## lm(formula = sqrt(IPT) ~ FCDo + FCDi + CC + CP + as.factor(DEPENDENCIAADM) +
## as.factor(CONCEITO) + as.factor(UFPROGRAMA) + FCDi * CC -
## 1, data = IMI_e_IPT)
##
## Residuals:
## Min 1Q Median 3Q Max
## -0.84317 -0.26218 -0.09848 0.17089 1.83521
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## FCDo 0.296810 0.045703 6.494 9.82e-11 ***
## FCDi 0.035470 0.114392 0.310 0.756529
## CC -0.305536 0.093859 -3.255 0.001146 **
## CP 0.078916 0.066954 1.179 0.238639
## as.factor(DEPENDENCIAADM)PRIVADA 0.005782 0.214983 0.027 0.978545
## as.factor(DEPENDENCIAADM)PÚBLICA -0.052682 0.213397 -0.247 0.805023
## as.factor(CONCEITO)4 0.051471 0.017363 2.964 0.003059 **
## as.factor(CONCEITO)5 0.078061 0.023378 3.339 0.000851 ***
## as.factor(CONCEITO)6 0.143020 0.030833 4.638 3.67e-06 ***
## as.factor(CONCEITO)7 0.162759 0.040396 4.029 5.75e-05 ***
## as.factor(UFPROGRAMA)AL 0.336929 0.212819 1.583 0.113495
## as.factor(UFPROGRAMA)AM 0.170054 0.210222 0.809 0.418627
## as.factor(UFPROGRAMA)AP 0.286551 0.286711 0.999 0.317666
## as.factor(UFPROGRAMA)BA 0.251209 0.204912 1.226 0.220324
## as.factor(UFPROGRAMA)CE 0.244529 0.206071 1.187 0.235475
## as.factor(UFPROGRAMA)DF 0.237316 0.206512 1.149 0.250585
## as.factor(UFPROGRAMA)ES 0.195708 0.208323 0.939 0.347582
## as.factor(UFPROGRAMA)GO 0.200451 0.206537 0.971 0.331863
## as.factor(UFPROGRAMA)MA 0.205383 0.209770 0.979 0.327620
## as.factor(UFPROGRAMA)MG 0.287319 0.203846 1.409 0.158799
## as.factor(UFPROGRAMA)MS 0.178421 0.210932 0.846 0.397698
## as.factor(UFPROGRAMA)MT 0.147492 0.208995 0.706 0.480420
## as.factor(UFPROGRAMA)PA 0.245011 0.206412 1.187 0.235327
## as.factor(UFPROGRAMA)PB 0.230307 0.213089 1.081 0.279877
## as.factor(UFPROGRAMA)PE 0.291358 0.205269 1.419 0.155894
## as.factor(UFPROGRAMA)PI 0.356397 0.231281 1.541 0.123436
## as.factor(UFPROGRAMA)PR 0.291522 0.204107 1.428 0.153322
## as.factor(UFPROGRAMA)RJ 0.204828 0.203971 1.004 0.315368
## as.factor(UFPROGRAMA)RN 0.333091 0.206613 1.612 0.107042
## as.factor(UFPROGRAMA)RO 0.228228 0.231169 0.987 0.323591
## as.factor(UFPROGRAMA)RR -0.055025 0.405356 -0.136 0.892033
## as.factor(UFPROGRAMA)RS 0.268258 0.204146 1.314 0.188937
## as.factor(UFPROGRAMA)SC 0.230534 0.205866 1.120 0.262883
## as.factor(UFPROGRAMA)SE 0.717449 0.225157 3.186 0.001456 **
## as.factor(UFPROGRAMA)SP 0.186287 0.203858 0.914 0.360896
## as.factor(UFPROGRAMA)TO 0.275421 0.217391 1.267 0.205282
## FCDi:CC 0.083228 0.177392 0.469 0.638980
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.351 on 2821 degrees of freedom
## Multiple R-squared: 0.4233, Adjusted R-squared: 0.4157
## F-statistic: 55.96 on 37 and 2821 DF, p-value: < 2.2e-16## Warning: not plotting observations with leverage one:
## 1268
## Warning: not plotting observations with leverage one:
## 1268
Não possui normalidade nos dados para garantir os pressupostos da análise
1.4.1.2 Regressão de Poisson
regpoisson=glm(sqrt(IPT) ~ FCDo + FCDi + CC + CP + as.factor(DEPENDENCIAADM) + as.factor(CONCEITO) + as.factor(UFPROGRAMA) + FCDi * CC -1, family="poisson", data=IMI_e_IPT)
summary(regpoisson)##
## Call:
## glm(formula = sqrt(IPT) ~ FCDo + FCDi + CC + CP + as.factor(DEPENDENCIAADM) +
## as.factor(CONCEITO) + as.factor(UFPROGRAMA) + FCDi * CC -
## 1, family = "poisson", data = IMI_e_IPT)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -1.4073 -0.7081 -0.2271 0.3074 2.1756
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## FCDo 1.073340 0.245773 4.367 1.26e-05 ***
## FCDi 0.008651 0.642063 0.013 0.98925
## CC -1.359096 0.565538 -2.403 0.01625 *
## CP 0.292796 0.346992 0.844 0.39877
## as.factor(DEPENDENCIAADM)PRIVADA -14.173973 440.683395 -0.032 0.97434
## as.factor(DEPENDENCIAADM)PÚBLICA -14.363336 440.683374 -0.033 0.97400
## as.factor(CONCEITO)4 0.181913 0.093508 1.945 0.05173 .
## as.factor(CONCEITO)5 0.265754 0.123789 2.147 0.03181 *
## as.factor(CONCEITO)6 0.473919 0.154755 3.062 0.00220 **
## as.factor(CONCEITO)7 0.562573 0.204935 2.745 0.00605 **
## as.factor(UFPROGRAMA)AL 13.178131 440.683337 0.030 0.97614
## as.factor(UFPROGRAMA)AM 12.529444 440.683384 0.028 0.97732
## as.factor(UFPROGRAMA)AP 13.029627 440.684315 0.030 0.97641
## as.factor(UFPROGRAMA)BA 12.908784 440.683253 0.029 0.97663
## as.factor(UFPROGRAMA)CE 12.876676 440.683271 0.029 0.97669
## as.factor(UFPROGRAMA)DF 12.861083 440.683272 0.029 0.97672
## as.factor(UFPROGRAMA)ES 12.683449 440.683319 0.029 0.97704
## as.factor(UFPROGRAMA)GO 12.689389 440.683291 0.029 0.97703
## as.factor(UFPROGRAMA)MA 12.705164 440.683340 0.029 0.97700
## as.factor(UFPROGRAMA)MG 13.019330 440.683237 0.030 0.97643
## as.factor(UFPROGRAMA)MS 12.574788 440.683379 0.029 0.97724
## as.factor(UFPROGRAMA)MT 12.389935 440.683374 0.028 0.97757
## as.factor(UFPROGRAMA)PA 12.887794 440.683276 0.029 0.97667
## as.factor(UFPROGRAMA)PB 12.827642 440.683387 0.029 0.97678
## as.factor(UFPROGRAMA)PE 13.039558 440.683254 0.030 0.97639
## as.factor(UFPROGRAMA)PI 13.255723 440.683543 0.030 0.97600
## as.factor(UFPROGRAMA)PR 13.038930 440.683240 0.030 0.97640
## as.factor(UFPROGRAMA)RJ 12.740610 440.683242 0.029 0.97694
## as.factor(UFPROGRAMA)RN 13.173844 440.683265 0.030 0.97615
## as.factor(UFPROGRAMA)RO 12.799823 440.683760 0.029 0.97683
## as.factor(UFPROGRAMA)RR -0.201781 890.473538 0.000 0.99982
## as.factor(UFPROGRAMA)RS 12.955616 440.683241 0.029 0.97655
## as.factor(UFPROGRAMA)SC 12.835612 440.683264 0.029 0.97676
## as.factor(UFPROGRAMA)SE 13.863189 440.683341 0.031 0.97490
## as.factor(UFPROGRAMA)SP 12.673996 440.683240 0.029 0.97706
## as.factor(UFPROGRAMA)TO 13.004187 440.683410 0.030 0.97646
## FCDi:CC 0.542113 1.045530 0.519 0.60411
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for poisson family taken to be 1)
##
## Null deviance: 3359.1 on 2858 degrees of freedom
## Residual deviance: 1226.9 on 2821 degrees of freedom
## AIC: Inf
##
## Number of Fisher Scoring iterations: 121.4.1.3 Árvore de decisão
library(rpart)
library(rpart.plot)
arvore <- rpart(sqrt(IPT) ~ FCDo + FCDi + CC + CP + as.factor(DEPENDENCIAADM) + as.factor(CONCEITO), data = IMI_e_IPT)
summary(arvore)## Call:
## rpart(formula = sqrt(IPT) ~ FCDo + FCDi + CC + CP + as.factor(DEPENDENCIAADM) +
## as.factor(CONCEITO), data = IMI_e_IPT)
## n= 2858
##
## CP nsplit rel error xerror xstd
## 1 0.01420417 0 1.0000000 1.0004892 0.03933496
## 2 0.01060735 2 0.9715917 0.9823271 0.03806442
## 3 0.01000000 3 0.9609843 0.9814367 0.03816198
##
## Variable importance
## FCDo CC FCDi
## 50 23 14
## as.factor(CONCEITO) as.factor(DEPENDENCIAADM) CP
## 7 3 2
##
## Node number 1: 2858 observations, complexity param=0.01420417
## mean=0.2842771, MSE=0.1300202
## left son=2 (1465 obs) right son=3 (1393 obs)
## Primary splits:
## CC < 0.5714653 to the right, improve=0.012976080, (0 missing)
## FCDo < 0.5373357 to the left, improve=0.012306860, (0 missing)
## as.factor(DEPENDENCIAADM) splits as RL, improve=0.004831066, (0 missing)
## as.factor(CONCEITO) splits as LRRRR, improve=0.004750481, (0 missing)
## CP < 0.4605808 to the left, improve=0.003011676, (0 missing)
## Surrogate splits:
## FCDi < 0.3727152 to the right, agree=0.687, adj=0.358, (0 split)
## FCDo < 0.4293395 to the right, agree=0.648, adj=0.277, (0 split)
## as.factor(CONCEITO) splits as LRRRR, agree=0.642, adj=0.266, (0 split)
## as.factor(DEPENDENCIAADM) splits as LR, agree=0.537, adj=0.050, (0 split)
## CP < 0.5328175 to the left, agree=0.526, adj=0.027, (0 split)
##
## Node number 2: 1465 observations, complexity param=0.01060735
## mean=0.2442242, MSE=0.1189336
## left son=4 (771 obs) right son=5 (694 obs)
## Primary splits:
## FCDo < 0.570028 to the left, improve=0.022622340, (0 missing)
## as.factor(DEPENDENCIAADM) splits as RL, improve=0.012423700, (0 missing)
## FCDi < 0.4847926 to the left, improve=0.008504852, (0 missing)
## as.factor(CONCEITO) splits as LRRRL, improve=0.006687891, (0 missing)
## CC < 1.022328 to the right, improve=0.006214590, (0 missing)
## Surrogate splits:
## FCDi < 0.4644661 to the left, agree=0.655, adj=0.271, (0 split)
## as.factor(DEPENDENCIAADM) splits as RL, agree=0.579, adj=0.111, (0 split)
## CP < 0.4581944 to the left, agree=0.576, adj=0.105, (0 split)
## CC < 0.6402778 to the left, agree=0.560, adj=0.072, (0 split)
## as.factor(CONCEITO) splits as RLLLL, agree=0.552, adj=0.055, (0 split)
##
## Node number 3: 1393 observations, complexity param=0.01420417
## mean=0.3264003, MSE=0.1382183
## left son=6 (1039 obs) right son=7 (354 obs)
## Primary splits:
## FCDo < 0.5373357 to the left, improve=0.029784190, (0 missing)
## CP < 0.4523575 to the left, improve=0.005847186, (0 missing)
## FCDi < 0.5564663 to the left, improve=0.004290214, (0 missing)
## as.factor(DEPENDENCIAADM) splits as RL, improve=0.002849396, (0 missing)
## as.factor(CONCEITO) splits as LLLRR, improve=0.002368822, (0 missing)
## Surrogate splits:
## FCDi < 0.7564272 to the left, agree=0.753, adj=0.028, (0 split)
## CC < 0.5708403 to the left, agree=0.747, adj=0.003, (0 split)
##
## Node number 4: 771 observations
## mean=0.1950119, MSE=0.09876409
##
## Node number 5: 694 observations
## mean=0.2988967, MSE=0.1356614
##
## Node number 6: 1039 observations
## mean=0.2889487, MSE=0.1104584
##
## Node number 7: 354 observations
## mean=0.4363215, MSE=0.203495
1.4.2 Sem outliers
fitIPT<- lm(formula = sqrt(IPT) ~ FCDo + FCDi + CC + CP + as.factor(DEPENDENCIAADM) + as.factor(CONCEITO) + as.factor(UFPROGRAMA) - 1, data = no_outliers)
summary(fitIPT)##
## Call:
## lm(formula = sqrt(IPT) ~ FCDo + FCDi + CC + CP + as.factor(DEPENDENCIAADM) +
## as.factor(CONCEITO) + as.factor(UFPROGRAMA) - 1, data = no_outliers)
##
## Residuals:
## Min 1Q Median 3Q Max
## -0.48865 -0.17996 -0.05381 0.13491 1.14266
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## FCDo -0.032993 0.107500 -0.307 0.7591
## FCDi 0.087610 0.091725 0.955 0.3402
## CC 0.029715 0.099763 0.298 0.7660
## CP 0.332823 0.160761 2.070 0.0392 *
## as.factor(DEPENDENCIAADM)PRIVADA 0.886206 0.148588 5.964 6.60e-09 ***
## as.factor(DEPENDENCIAADM)PÚBLICA 0.823164 0.138450 5.946 7.31e-09 ***
## as.factor(CONCEITO)4 0.061117 0.040729 1.501 0.1345
## as.factor(CONCEITO)5 0.005148 0.051596 0.100 0.9206
## as.factor(CONCEITO)6 0.067716 0.063203 1.071 0.2848
## as.factor(CONCEITO)7 0.117540 0.089545 1.313 0.1903
## as.factor(UFPROGRAMA)AM 0.004078 0.189628 0.022 0.9829
## as.factor(UFPROGRAMA)AP -0.221887 0.293099 -0.757 0.4496
## as.factor(UFPROGRAMA)BA -0.068495 0.123099 -0.556 0.5783
## as.factor(UFPROGRAMA)CE -0.102848 0.137101 -0.750 0.4537
## as.factor(UFPROGRAMA)DF -0.163843 0.128162 -1.278 0.2021
## as.factor(UFPROGRAMA)ES -0.065050 0.171513 -0.379 0.7047
## as.factor(UFPROGRAMA)GO -0.114053 0.147563 -0.773 0.4402
## as.factor(UFPROGRAMA)MA 0.086344 0.172345 0.501 0.6167
## as.factor(UFPROGRAMA)MG -0.082697 0.113330 -0.730 0.4661
## as.factor(UFPROGRAMA)MS -0.302045 0.195083 -1.548 0.1226
## as.factor(UFPROGRAMA)MT 0.023450 0.222787 0.105 0.9162
## as.factor(UFPROGRAMA)PA -0.048164 0.160980 -0.299 0.7650
## as.factor(UFPROGRAMA)PB -0.052877 0.172490 -0.307 0.7594
## as.factor(UFPROGRAMA)PE 0.021935 0.121647 0.180 0.8570
## as.factor(UFPROGRAMA)PI -0.135106 0.188768 -0.716 0.4747
## as.factor(UFPROGRAMA)PR 0.007321 0.115117 0.064 0.9493
## as.factor(UFPROGRAMA)RJ -0.056194 0.116527 -0.482 0.6300
## as.factor(UFPROGRAMA)RN 0.001444 0.125321 0.012 0.9908
## as.factor(UFPROGRAMA)RO -0.290385 0.293091 -0.991 0.3226
## as.factor(UFPROGRAMA)RS 0.079893 0.114530 0.698 0.4860
## as.factor(UFPROGRAMA)SC -0.091232 0.128947 -0.708 0.4798
## as.factor(UFPROGRAMA)SE 0.316064 0.149289 2.117 0.0350 *
## as.factor(UFPROGRAMA)SP -0.101283 0.117108 -0.865 0.3878
## as.factor(UFPROGRAMA)TO -0.186034 0.220446 -0.844 0.3994
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.2727 on 315 degrees of freedom
## Multiple R-squared: 0.9412, Adjusted R-squared: 0.9348
## F-statistic: 148.2 on 34 and 315 DF, p-value: < 2.2e-16## Warning: not plotting observations with leverage one:
## 97, 106
## Warning: not plotting observations with leverage one:
## 97, 106
Próximo da normalidade mas ainda não consegue garantir os pressupostos
## [1] "Correlação normalizada 0.0929039972121071"## [1] "Correlação sem normalização -0.000893884909316986"1.4.2.1 Regressão de Poisson
regpoisson=glm(sqrt(IPT) ~ FCDo + FCDi + CC + CP + as.factor(DEPENDENCIAADM) + as.factor(CONCEITO) + as.factor(UFPROGRAMA) + FCDi * CC -1, family="poisson", data= IMI_e_IPT)
summary(regpoisson)##
## Call:
## glm(formula = sqrt(IPT) ~ FCDo + FCDi + CC + CP + as.factor(DEPENDENCIAADM) +
## as.factor(CONCEITO) + as.factor(UFPROGRAMA) + FCDi * CC -
## 1, family = "poisson", data = IMI_e_IPT)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -1.4073 -0.7081 -0.2271 0.3074 2.1756
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## FCDo 1.073340 0.245773 4.367 1.26e-05 ***
## FCDi 0.008651 0.642063 0.013 0.98925
## CC -1.359096 0.565538 -2.403 0.01625 *
## CP 0.292796 0.346992 0.844 0.39877
## as.factor(DEPENDENCIAADM)PRIVADA -14.173973 440.683395 -0.032 0.97434
## as.factor(DEPENDENCIAADM)PÚBLICA -14.363336 440.683374 -0.033 0.97400
## as.factor(CONCEITO)4 0.181913 0.093508 1.945 0.05173 .
## as.factor(CONCEITO)5 0.265754 0.123789 2.147 0.03181 *
## as.factor(CONCEITO)6 0.473919 0.154755 3.062 0.00220 **
## as.factor(CONCEITO)7 0.562573 0.204935 2.745 0.00605 **
## as.factor(UFPROGRAMA)AL 13.178131 440.683337 0.030 0.97614
## as.factor(UFPROGRAMA)AM 12.529444 440.683384 0.028 0.97732
## as.factor(UFPROGRAMA)AP 13.029627 440.684315 0.030 0.97641
## as.factor(UFPROGRAMA)BA 12.908784 440.683253 0.029 0.97663
## as.factor(UFPROGRAMA)CE 12.876676 440.683271 0.029 0.97669
## as.factor(UFPROGRAMA)DF 12.861083 440.683272 0.029 0.97672
## as.factor(UFPROGRAMA)ES 12.683449 440.683319 0.029 0.97704
## as.factor(UFPROGRAMA)GO 12.689389 440.683291 0.029 0.97703
## as.factor(UFPROGRAMA)MA 12.705164 440.683340 0.029 0.97700
## as.factor(UFPROGRAMA)MG 13.019330 440.683237 0.030 0.97643
## as.factor(UFPROGRAMA)MS 12.574788 440.683379 0.029 0.97724
## as.factor(UFPROGRAMA)MT 12.389935 440.683374 0.028 0.97757
## as.factor(UFPROGRAMA)PA 12.887794 440.683276 0.029 0.97667
## as.factor(UFPROGRAMA)PB 12.827642 440.683387 0.029 0.97678
## as.factor(UFPROGRAMA)PE 13.039558 440.683254 0.030 0.97639
## as.factor(UFPROGRAMA)PI 13.255723 440.683543 0.030 0.97600
## as.factor(UFPROGRAMA)PR 13.038930 440.683240 0.030 0.97640
## as.factor(UFPROGRAMA)RJ 12.740610 440.683242 0.029 0.97694
## as.factor(UFPROGRAMA)RN 13.173844 440.683265 0.030 0.97615
## as.factor(UFPROGRAMA)RO 12.799823 440.683760 0.029 0.97683
## as.factor(UFPROGRAMA)RR -0.201781 890.473538 0.000 0.99982
## as.factor(UFPROGRAMA)RS 12.955616 440.683241 0.029 0.97655
## as.factor(UFPROGRAMA)SC 12.835612 440.683264 0.029 0.97676
## as.factor(UFPROGRAMA)SE 13.863189 440.683341 0.031 0.97490
## as.factor(UFPROGRAMA)SP 12.673996 440.683240 0.029 0.97706
## as.factor(UFPROGRAMA)TO 13.004187 440.683410 0.030 0.97646
## FCDi:CC 0.542113 1.045530 0.519 0.60411
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for poisson family taken to be 1)
##
## Null deviance: 3359.1 on 2858 degrees of freedom
## Residual deviance: 1226.9 on 2821 degrees of freedom
## AIC: Inf
##
## Number of Fisher Scoring iterations: 121.4.2.2 Árvore de decisão
arvore_nout <- rpart(sqrt(IPT) ~ FCDo + FCDi + CC + CP + as.factor(DEPENDENCIAADM) + as.factor(CONCEITO), data = no_outliers)
summary(arvore_nout)## Call:
## rpart(formula = sqrt(IPT) ~ FCDo + FCDi + CC + CP + as.factor(DEPENDENCIAADM) +
## as.factor(CONCEITO), data = no_outliers)
## n= 349
##
## CP nsplit rel error xerror xstd
## 1 0.02282279 0 1.0000000 1.005645 0.1197302
## 2 0.01774939 3 0.9294627 1.152528 0.1290911
## 3 0.01624681 8 0.8385899 1.226711 0.1335846
## 4 0.01436838 9 0.8223431 1.245826 0.1332226
## 5 0.01114782 10 0.8079747 1.290706 0.1375681
## 6 0.01084124 14 0.7633834 1.329958 0.1431779
## 7 0.01020310 16 0.7417010 1.328174 0.1432480
## 8 0.01000000 18 0.7212948 1.325013 0.1430175
##
## Variable importance
## FCDi FCDo CC
## 30 21 20
## as.factor(CONCEITO) CP as.factor(DEPENDENCIAADM)
## 15 9 5
##
## Node number 1: 349 observations, complexity param=0.02282279
## mean=1.03088, MSE=0.07822102
## left son=2 (270 obs) right son=3 (79 obs)
## Primary splits:
## as.factor(DEPENDENCIAADM) splits as RL, improve=0.018491400, (0 missing)
## CP < 0.6217262 to the left, improve=0.016534390, (0 missing)
## CC < 0.9772398 to the left, improve=0.016472820, (0 missing)
## FCDi < 0.9089069 to the left, improve=0.011495880, (0 missing)
## FCDo < 0.5055856 to the left, improve=0.009905112, (0 missing)
## Surrogate splits:
## FCDo < 0.7656863 to the left, agree=0.805, adj=0.139, (0 split)
## FCDi < 0.8198052 to the left, agree=0.785, adj=0.051, (0 split)
## CP < 0.2847222 to the right, agree=0.779, adj=0.025, (0 split)
##
## Node number 2: 270 observations, complexity param=0.01774939
## mean=1.010308, MSE=0.06725982
## left son=4 (9 obs) right son=5 (261 obs)
## Primary splits:
## FCDi < 0.8153409 to the right, improve=0.018277410, (0 missing)
## CP < 0.4491571 to the left, improve=0.017833270, (0 missing)
## FCDo < 0.5055856 to the left, improve=0.012773410, (0 missing)
## CC < 0.5241911 to the left, improve=0.012177220, (0 missing)
## as.factor(CONCEITO) splits as LLLRR, improve=0.005276337, (0 missing)
##
## Node number 3: 79 observations, complexity param=0.02282279
## mean=1.101189, MSE=0.1092935
## left son=6 (72 obs) right son=7 (7 obs)
## Primary splits:
## FCDi < 0.8568548 to the left, improve=0.08585471, (0 missing)
## as.factor(CONCEITO) splits as LRLLL, improve=0.05595841, (0 missing)
## CP < 0.5502976 to the left, improve=0.04253452, (0 missing)
## CC < 0.744621 to the left, improve=0.02808771, (0 missing)
## FCDo < 0.5505051 to the right, improve=0.02545831, (0 missing)
##
## Node number 4: 9 observations
## mean=0.8214936, MSE=0.01130151
##
## Node number 5: 261 observations, complexity param=0.01774939
## mean=1.016818, MSE=0.06791769
## left son=10 (110 obs) right son=11 (151 obs)
## Primary splits:
## CC < 0.5241911 to the left, improve=0.017179430, (0 missing)
## FCDo < 0.5055856 to the left, improve=0.016499930, (0 missing)
## CP < 0.4491571 to the left, improve=0.013661420, (0 missing)
## FCDi < 0.305839 to the left, improve=0.009885776, (0 missing)
## as.factor(CONCEITO) splits as LLLRR, improve=0.003872234, (0 missing)
## Surrogate splits:
## FCDi < 0.2837121 to the left, agree=0.759, adj=0.427, (0 split)
## as.factor(CONCEITO) splits as RRLLL, agree=0.655, adj=0.182, (0 split)
## FCDo < 0.4359649 to the left, agree=0.651, adj=0.173, (0 split)
## CP < 0.3026042 to the left, agree=0.586, adj=0.018, (0 split)
##
## Node number 6: 72 observations, complexity param=0.02282279
## mean=1.070985, MSE=0.08319642
## left son=12 (44 obs) right son=13 (28 obs)
## Primary splits:
## as.factor(CONCEITO) splits as LRLLL, improve=0.11344000, (0 missing)
## CP < 0.5502976 to the left, improve=0.10829070, (0 missing)
## FCDo < 0.4742063 to the right, improve=0.03983165, (0 missing)
## CC < 0.4436264 to the right, improve=0.03289703, (0 missing)
## FCDi < 0.5931034 to the right, improve=0.03242660, (0 missing)
## Surrogate splits:
## FCDo < 0.4880952 to the right, agree=0.722, adj=0.286, (0 split)
## CC < 0.4681691 to the right, agree=0.653, adj=0.107, (0 split)
## FCDi < 0.2265512 to the right, agree=0.625, adj=0.036, (0 split)
## CP < 0.5390857 to the left, agree=0.625, adj=0.036, (0 split)
##
## Node number 7: 7 observations
## mean=1.411857, MSE=0.2718225
##
## Node number 10: 110 observations, complexity param=0.01436838
## mean=0.9767974, MSE=0.06031334
## left son=20 (99 obs) right son=21 (11 obs)
## Primary splits:
## FCDi < 0.5776093 to the left, improve=0.05912221, (0 missing)
## CC < 0.3458333 to the right, improve=0.03839310, (0 missing)
## as.factor(CONCEITO) splits as LLLLR, improve=0.02705273, (0 missing)
## CP < 0.4422619 to the left, improve=0.01626740, (0 missing)
## FCDo < 0.3181515 to the left, improve=0.01159944, (0 missing)
## Surrogate splits:
## FCDo < 0.6777778 to the left, agree=0.909, adj=0.091, (0 split)
##
## Node number 11: 151 observations, complexity param=0.01774939
## mean=1.045973, MSE=0.07144051
## left son=22 (78 obs) right son=23 (73 obs)
## Primary splits:
## FCDi < 0.5116238 to the right, improve=0.049506290, (0 missing)
## CC < 0.5288521 to the right, improve=0.047210890, (0 missing)
## CP < 0.4922794 to the left, improve=0.035483860, (0 missing)
## FCDo < 0.7064327 to the right, improve=0.013602440, (0 missing)
## as.factor(CONCEITO) splits as LLLRL, improve=0.008312707, (0 missing)
## Surrogate splits:
## as.factor(CONCEITO) splits as LRRRR, agree=0.709, adj=0.397, (0 split)
## FCDo < 0.53125 to the right, agree=0.689, adj=0.356, (0 split)
## CC < 0.6512144 to the right, agree=0.649, adj=0.274, (0 split)
## CP < 0.5114426 to the left, agree=0.563, adj=0.096, (0 split)
##
## Node number 12: 44 observations, complexity param=0.0102031
## mean=0.9934879, MSE=0.06276538
## left son=24 (35 obs) right son=25 (9 obs)
## Primary splits:
## CC < 0.5134056 to the right, improve=0.07949447, (0 missing)
## FCDi < 0.5528821 to the left, improve=0.07729982, (0 missing)
## CP < 0.4094907 to the left, improve=0.07376566, (0 missing)
## FCDo < 0.8176638 to the right, improve=0.03895006, (0 missing)
## as.factor(CONCEITO) splits as R-LLL, improve=0.01075653, (0 missing)
##
## Node number 13: 28 observations, complexity param=0.01624681
## mean=1.192767, MSE=0.09103369
## left son=26 (10 obs) right son=27 (18 obs)
## Primary splits:
## FCDi < 0.5760369 to the right, improve=0.17400310, (0 missing)
## FCDo < 0.4742063 to the right, improve=0.08989864, (0 missing)
## CP < 0.4949875 to the left, improve=0.08841769, (0 missing)
## CC < 0.6364286 to the right, improve=0.08680351, (0 missing)
## Surrogate splits:
## FCDo < 0.6237374 to the right, agree=0.679, adj=0.1, (0 split)
##
## Node number 20: 99 observations, complexity param=0.01114782
## mean=0.9568925, MSE=0.04303043
## left son=40 (88 obs) right son=41 (11 obs)
## Primary splits:
## as.factor(CONCEITO) splits as LLLLR, improve=0.05782292, (0 missing)
## FCDo < 0.6262255 to the right, improve=0.04835172, (0 missing)
## CC < 0.4221709 to the left, improve=0.03187391, (0 missing)
## FCDi < 0.3762626 to the right, improve=0.02416534, (0 missing)
## CP < 0.3844157 to the right, improve=0.01702926, (0 missing)
##
## Node number 21: 11 observations
## mean=1.155942, MSE=0.180201
##
## Node number 22: 78 observations, complexity param=0.01084124
## mean=0.9884398, MSE=0.04855783
## left son=44 (53 obs) right son=45 (25 obs)
## Primary splits:
## FCDo < 0.5555556 to the right, improve=0.05771495, (0 missing)
## CP < 0.3622024 to the right, improve=0.04915763, (0 missing)
## FCDi < 0.6236111 to the right, improve=0.03517085, (0 missing)
## as.factor(CONCEITO) splits as LRRL-, improve=0.02845517, (0 missing)
## CC < 0.5445374 to the right, improve=0.02243028, (0 missing)
## Surrogate splits:
## CC < 0.571131 to the right, agree=0.731, adj=0.16, (0 split)
## CP < 0.3479701 to the right, agree=0.692, adj=0.04, (0 split)
##
## Node number 23: 73 observations, complexity param=0.01774939
## mean=1.107446, MSE=0.08857475
## left son=46 (42 obs) right son=47 (31 obs)
## Primary splits:
## FCDo < 0.5196759 to the left, improve=0.11872500, (0 missing)
## FCDi < 0.4038198 to the left, improve=0.08744170, (0 missing)
## CC < 0.5299148 to the right, improve=0.07999502, (0 missing)
## CP < 0.493595 to the left, improve=0.06055149, (0 missing)
## as.factor(CONCEITO) splits as RLLRL, improve=0.04731229, (0 missing)
## Surrogate splits:
## FCDi < 0.3636541 to the left, agree=0.685, adj=0.258, (0 split)
## CP < 0.493595 to the left, agree=0.644, adj=0.161, (0 split)
## as.factor(CONCEITO) splits as RLLLL, agree=0.644, adj=0.161, (0 split)
## CC < 0.5254464 to the right, agree=0.603, adj=0.065, (0 split)
##
## Node number 24: 35 observations, complexity param=0.0102031
## mean=0.9576687, MSE=0.0545813
## left son=48 (22 obs) right son=49 (13 obs)
## Primary splits:
## CC < 0.7054784 to the left, improve=0.17668720, (0 missing)
## CP < 0.5079004 to the left, improve=0.10888740, (0 missing)
## FCDo < 0.5677656 to the left, improve=0.10299430, (0 missing)
## FCDi < 0.5278571 to the left, improve=0.09328107, (0 missing)
## as.factor(CONCEITO) splits as R-LLL, improve=0.01092535, (0 missing)
## Surrogate splits:
## CP < 0.2951389 to the right, agree=0.686, adj=0.154, (0 split)
## FCDo < 0.4444444 to the right, agree=0.657, adj=0.077, (0 split)
##
## Node number 25: 9 observations
## mean=1.132785, MSE=0.07019924
##
## Node number 26: 10 observations
## mean=1.023912, MSE=0.03782498
##
## Node number 27: 18 observations
## mean=1.286576, MSE=0.09595387
##
## Node number 40: 88 observations, complexity param=0.01114782
## mean=0.9392568, MSE=0.03773218
## left son=80 (8 obs) right son=81 (80 obs)
## Primary splits:
## FCDo < 0.6262255 to the right, improve=0.049166600, (0 missing)
## CC < 0.3982491 to the left, improve=0.034074560, (0 missing)
## CP < 0.3844157 to the right, improve=0.018987410, (0 missing)
## FCDi < 0.1707642 to the left, improve=0.018899550, (0 missing)
## as.factor(CONCEITO) splits as RRLL-, improve=0.005205054, (0 missing)
##
## Node number 41: 11 observations
## mean=1.097978, MSE=0.06302311
##
## Node number 44: 53 observations
## mean=0.9520813, MSE=0.02939089
##
## Node number 45: 25 observations, complexity param=0.01084124
## mean=1.06552, MSE=0.08044789
## left son=90 (18 obs) right son=91 (7 obs)
## Primary splits:
## FCDo < 0.5230856 to the left, improve=0.18561910, (0 missing)
## CC < 0.577178 to the left, improve=0.07196665, (0 missing)
## as.factor(CONCEITO) splits as RLRL-, improve=0.07046449, (0 missing)
## FCDi < 0.591253 to the left, improve=0.06953301, (0 missing)
## CP < 0.399213 to the right, improve=0.06489596, (0 missing)
## Surrogate splits:
## FCDi < 0.5732143 to the right, agree=0.76, adj=0.143, (0 split)
## CC < 0.5436709 to the right, agree=0.76, adj=0.143, (0 split)
## CP < 0.4987637 to the left, agree=0.76, adj=0.143, (0 split)
##
## Node number 46: 42 observations
## mean=1.019345, MSE=0.0388596
##
## Node number 47: 31 observations, complexity param=0.01774939
## mean=1.226809, MSE=0.1311672
## left son=94 (15 obs) right son=95 (16 obs)
## Primary splits:
## CP < 0.4890941 to the left, improve=0.13343660, (0 missing)
## FCDi < 0.4038198 to the left, improve=0.09019198, (0 missing)
## as.factor(CONCEITO) splits as RLLR-, improve=0.07319815, (0 missing)
## CC < 0.5389888 to the right, improve=0.06864687, (0 missing)
## FCDo < 0.6306043 to the right, improve=0.04451223, (0 missing)
## Surrogate splits:
## as.factor(CONCEITO) splits as RLRL-, agree=0.742, adj=0.467, (0 split)
## FCDi < 0.390628 to the left, agree=0.677, adj=0.333, (0 split)
## CC < 0.6080688 to the right, agree=0.581, adj=0.133, (0 split)
## FCDo < 0.5342262 to the right, agree=0.548, adj=0.067, (0 split)
##
## Node number 48: 22 observations
## mean=0.8821795, MSE=0.03318247
##
## Node number 49: 13 observations
## mean=1.08542, MSE=0.0648306
##
## Node number 80: 8 observations
## mean=0.8030524, MSE=0.002435638
##
## Node number 81: 80 observations, complexity param=0.01114782
## mean=0.9528772, MSE=0.03922116
## left son=162 (29 obs) right son=163 (51 obs)
## Primary splits:
## CC < 0.4221709 to the left, improve=0.047450190, (0 missing)
## FCDi < 0.1707642 to the left, improve=0.028815950, (0 missing)
## CP < 0.4306469 to the left, improve=0.021127690, (0 missing)
## FCDo < 0.5033333 to the left, improve=0.013459220, (0 missing)
## as.factor(CONCEITO) splits as LRLL-, improve=0.008112287, (0 missing)
## Surrogate splits:
## FCDo < 0.2472222 to the left, agree=0.675, adj=0.103, (0 split)
## FCDi < 0.1084967 to the left, agree=0.675, adj=0.103, (0 split)
## CP < 0.3364286 to the left, agree=0.650, adj=0.034, (0 split)
## as.factor(CONCEITO) splits as RRRL-, agree=0.650, adj=0.034, (0 split)
##
## Node number 90: 18 observations
## mean=0.9893152, MSE=0.05215333
##
## Node number 91: 7 observations
## mean=1.261475, MSE=0.09987436
##
## Node number 94: 15 observations
## mean=1.090174, MSE=0.05014305
##
## Node number 95: 16 observations
## mean=1.354905, MSE=0.1732162
##
## Node number 162: 29 observations
## mean=0.8956681, MSE=0.02279377
##
## Node number 163: 51 observations, complexity param=0.01114782
## mean=0.9854079, MSE=0.04564293
## left son=326 (44 obs) right son=327 (7 obs)
## Primary splits:
## CC < 0.4392778 to the right, improve=0.28303160, (0 missing)
## FCDi < 0.3762626 to the right, improve=0.05658093, (0 missing)
## CP < 0.3914764 to the right, improve=0.02654500, (0 missing)
## FCDo < 0.5175926 to the right, improve=0.01990692, (0 missing)
## as.factor(CONCEITO) splits as LRRL-, improve=0.01343994, (0 missing)
##
## Node number 326: 44 observations
## mean=0.9400736, MSE=0.02508037
##
## Node number 327: 7 observations
## mean=1.270366, MSE=0.08077355
1.5 Patentes
fitPatente<- lm(formula = sqrt(SPPP/QPPP) ~ FCDo + FCDi + CC + CP + as.factor(DEPENDENCIAADM) + as.factor(CONCEITO) + as.factor(UFPROGRAMA) + FCDi * CC - 1, data = no_outliers)
summary(fitPatente)##
## Call:
## lm(formula = sqrt(SPPP/QPPP) ~ FCDo + FCDi + CC + CP + as.factor(DEPENDENCIAADM) +
## as.factor(CONCEITO) + as.factor(UFPROGRAMA) + FCDi * CC -
## 1, data = no_outliers)
##
## Residuals:
## Min 1Q Median 3Q Max
## -1.04262 -0.22803 0.00558 0.20992 1.50475
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## FCDo 0.11819 0.15310 0.772 0.440695
## FCDi 0.47621 0.35180 1.354 0.176827
## CC -0.17147 0.30760 -0.557 0.577622
## CP 0.07917 0.22872 0.346 0.729457
## as.factor(DEPENDENCIAADM)PRIVADA 0.58341 0.25823 2.259 0.024551 *
## as.factor(DEPENDENCIAADM)PÚBLICA 0.60106 0.24599 2.443 0.015100 *
## as.factor(CONCEITO)4 0.15022 0.05795 2.592 0.009978 **
## as.factor(CONCEITO)5 0.18623 0.07340 2.537 0.011660 *
## as.factor(CONCEITO)6 0.31017 0.09000 3.446 0.000646 ***
## as.factor(CONCEITO)7 0.48170 0.12816 3.759 0.000204 ***
## as.factor(UFPROGRAMA)AM -0.51719 0.26986 -1.917 0.056209 .
## as.factor(UFPROGRAMA)AP 0.04558 0.41682 0.109 0.912994
## as.factor(UFPROGRAMA)BA -0.06116 0.17511 -0.349 0.727132
## as.factor(UFPROGRAMA)CE -0.23068 0.19504 -1.183 0.237802
## as.factor(UFPROGRAMA)DF -0.17233 0.18229 -0.945 0.345197
## as.factor(UFPROGRAMA)ES -0.07956 0.24396 -0.326 0.744542
## as.factor(UFPROGRAMA)GO -0.05098 0.21036 -0.242 0.808679
## as.factor(UFPROGRAMA)MA -0.02890 0.24510 -0.118 0.906223
## as.factor(UFPROGRAMA)MG -0.16847 0.16121 -1.045 0.296810
## as.factor(UFPROGRAMA)MS -0.16157 0.27755 -0.582 0.560882
## as.factor(UFPROGRAMA)MT 0.27199 0.31745 0.857 0.392206
## as.factor(UFPROGRAMA)PA 0.06686 0.22893 0.292 0.770424
## as.factor(UFPROGRAMA)PB -0.18183 0.24554 -0.741 0.459532
## as.factor(UFPROGRAMA)PE 0.07924 0.17331 0.457 0.647826
## as.factor(UFPROGRAMA)PI -0.14271 0.26855 -0.531 0.595502
## as.factor(UFPROGRAMA)PR -0.04901 0.16373 -0.299 0.764872
## as.factor(UFPROGRAMA)RJ -0.39161 0.16574 -2.363 0.018743 *
## as.factor(UFPROGRAMA)RN 0.09459 0.17828 0.531 0.596092
## as.factor(UFPROGRAMA)RO -0.24508 0.41685 -0.588 0.557000
## as.factor(UFPROGRAMA)RS -0.17845 0.16294 -1.095 0.274254
## as.factor(UFPROGRAMA)SC -0.34405 0.18345 -1.875 0.061653 .
## as.factor(UFPROGRAMA)SE 0.32892 0.21231 1.549 0.122344
## as.factor(UFPROGRAMA)SP -0.16600 0.16654 -0.997 0.319661
## as.factor(UFPROGRAMA)TO -0.51461 0.31350 -1.642 0.101694
## FCDi:CC -0.48437 0.55907 -0.866 0.386942
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.3879 on 314 degrees of freedom
## Multiple R-squared: 0.785, Adjusted R-squared: 0.761
## F-statistic: 32.75 on 35 and 314 DF, p-value: < 2.2e-161.5.0.1 Regressão de Poisson
regpoisson=glm(sqrt(SPPP/QPPP) ~ FCDo + FCDi + CC + CP + as.factor(DEPENDENCIAADM) + as.factor(CONCEITO) + as.factor(UFPROGRAMA) + FCDi * CC -1, family="poisson", data= IMI_e_IPT)
summary(regpoisson)##
## Call:
## glm(formula = sqrt(SPPP/QPPP) ~ FCDo + FCDi + CC + CP + as.factor(DEPENDENCIAADM) +
## as.factor(CONCEITO) + as.factor(UFPROGRAMA) + FCDi * CC -
## 1, family = "poisson", data = IMI_e_IPT)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -1.2005 -0.5658 -0.4486 0.2305 2.5971
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## FCDo 1.2155 0.3275 3.711 0.000206 ***
## FCDi -0.2236 0.8551 -0.262 0.793690
## CC -2.6057 0.7764 -3.356 0.000790 ***
## CP 0.8352 0.4402 1.897 0.057779 .
## as.factor(DEPENDENCIAADM)PRIVADA -13.9911 429.4776 -0.033 0.974012
## as.factor(DEPENDENCIAADM)PÚBLICA -14.0189 429.4776 -0.033 0.973960
## as.factor(CONCEITO)4 0.2469 0.1278 1.932 0.053351 .
## as.factor(CONCEITO)5 0.4300 0.1638 2.625 0.008670 **
## as.factor(CONCEITO)6 0.7350 0.1982 3.709 0.000208 ***
## as.factor(CONCEITO)7 0.9958 0.2498 3.986 6.72e-05 ***
## as.factor(UFPROGRAMA)AL 12.6915 429.4775 0.030 0.976425
## as.factor(UFPROGRAMA)AM 11.8077 429.4776 0.027 0.978066
## as.factor(UFPROGRAMA)AP 12.8809 429.4785 0.030 0.976073
## as.factor(UFPROGRAMA)BA 12.3216 429.4774 0.029 0.977112
## as.factor(UFPROGRAMA)CE 12.2236 429.4774 0.028 0.977294
## as.factor(UFPROGRAMA)DF 12.0511 429.4774 0.028 0.977614
## as.factor(UFPROGRAMA)ES 12.1617 429.4775 0.028 0.977409
## as.factor(UFPROGRAMA)GO 12.0979 429.4775 0.028 0.977527
## as.factor(UFPROGRAMA)MA 12.2703 429.4775 0.029 0.977207
## as.factor(UFPROGRAMA)MG 12.3975 429.4774 0.029 0.976971
## as.factor(UFPROGRAMA)MS 12.1020 429.4776 0.028 0.977520
## as.factor(UFPROGRAMA)MT 12.0404 429.4775 0.028 0.977634
## as.factor(UFPROGRAMA)PA 12.1447 429.4774 0.028 0.977440
## as.factor(UFPROGRAMA)PB 12.3053 429.4776 0.029 0.977142
## as.factor(UFPROGRAMA)PE 12.6171 429.4774 0.029 0.976563
## as.factor(UFPROGRAMA)PI 12.8485 429.4778 0.030 0.976134
## as.factor(UFPROGRAMA)PR 12.4447 429.4774 0.029 0.976884
## as.factor(UFPROGRAMA)RJ 11.6693 429.4774 0.027 0.978323
## as.factor(UFPROGRAMA)RN 12.7194 429.4774 0.030 0.976373
## as.factor(UFPROGRAMA)RO 11.8032 429.4788 0.027 0.978075
## as.factor(UFPROGRAMA)RR -0.3080 884.9815 0.000 0.999722
## as.factor(UFPROGRAMA)RS 12.1394 429.4774 0.028 0.977450
## as.factor(UFPROGRAMA)SC 11.9645 429.4774 0.028 0.977775
## as.factor(UFPROGRAMA)SE 13.5189 429.4775 0.031 0.974889
## as.factor(UFPROGRAMA)SP 11.9685 429.4774 0.028 0.977768
## as.factor(UFPROGRAMA)TO 12.2209 429.4777 0.028 0.977299
## FCDi:CC 1.3029 1.4374 0.906 0.364703
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for poisson family taken to be 1)
##
## Null deviance: 4275.4 on 2858 degrees of freedom
## Residual deviance: 1079.5 on 2821 degrees of freedom
## AIC: Inf
##
## Number of Fisher Scoring iterations: 121.5.0.2 Árvore de decisão
arvore_patente <- rpart(SPPP/QPPP ~ FCDo + FCDi + CC + CP + as.factor(DEPENDENCIAADM) + as.factor(CONCEITO), data = no_outliers)
summary(arvore_patente)## Call:
## rpart(formula = SPPP/QPPP ~ FCDo + FCDi + CC + CP + as.factor(DEPENDENCIAADM) +
## as.factor(CONCEITO), data = no_outliers)
## n= 349
##
## CP nsplit rel error xerror xstd
## 1 0.03547831 0 1.0000000 1.007428 0.1963998
## 2 0.02239911 2 0.9290434 1.106991 0.1984569
## 3 0.01844940 5 0.8618460 1.144126 0.1978482
## 4 0.01239861 6 0.8433966 1.171633 0.2045524
## 5 0.01029784 7 0.8309980 1.186228 0.2054333
## 6 0.01000000 10 0.8001045 1.185827 0.2052137
##
## Variable importance
## FCDi FCDo CC
## 29 25 18
## CP as.factor(CONCEITO) as.factor(DEPENDENCIAADM)
## 15 12 1
##
## Node number 1: 349 observations, complexity param=0.03547831
## mean=0.629482, MSE=0.424635
## left son=2 (148 obs) right son=3 (201 obs)
## Primary splits:
## CP < 0.4476111 to the left, improve=0.03222377, (0 missing)
## as.factor(CONCEITO) splits as LLLRR, improve=0.03009496, (0 missing)
## CC < 0.3445861 to the right, improve=0.02901461, (0 missing)
## FCDo < 0.7421652 to the right, improve=0.01929228, (0 missing)
## FCDi < 0.4241932 to the right, improve=0.01206120, (0 missing)
## Surrogate splits:
## FCDo < 0.2593656 to the left, agree=0.605, adj=0.068, (0 split)
## FCDi < 0.1863082 to the left, agree=0.605, adj=0.068, (0 split)
## CC < 0.8354167 to the right, agree=0.605, adj=0.068, (0 split)
##
## Node number 2: 148 observations, complexity param=0.0184494
## mean=0.4931608, MSE=0.3104785
## left son=4 (141 obs) right son=5 (7 obs)
## Primary splits:
## FCDi < 0.9089069 to the left, improve=0.05950184, (0 missing)
## FCDo < 0.6794872 to the left, improve=0.03972066, (0 missing)
## CC < 0.8143056 to the left, improve=0.03121239, (0 missing)
## as.factor(CONCEITO) splits as LRLRR, improve=0.02188764, (0 missing)
## as.factor(DEPENDENCIAADM) splits as RL, improve=0.01734681, (0 missing)
##
## Node number 3: 201 observations, complexity param=0.03547831
## mean=0.7298578, MSE=0.484932
## left son=6 (29 obs) right son=7 (172 obs)
## Primary splits:
## FCDo < 0.7421652 to the right, improve=0.05889031, (0 missing)
## CC < 0.6693845 to the right, improve=0.05210645, (0 missing)
## as.factor(CONCEITO) splits as LLLRR, improve=0.04265311, (0 missing)
## FCDi < 0.343002 to the right, improve=0.03777498, (0 missing)
## as.factor(DEPENDENCIAADM) splits as LR, improve=0.03187693, (0 missing)
##
## Node number 4: 141 observations, complexity param=0.01239861
## mean=0.4628763, MSE=0.2099164
## left son=8 (76 obs) right son=9 (65 obs)
## Primary splits:
## as.factor(CONCEITO) splits as LRLRR, improve=0.06207956, (0 missing)
## CC < 0.5957221 to the right, improve=0.02193168, (0 missing)
## FCDi < 0.6492599 to the right, improve=0.01779479, (0 missing)
## FCDo < 0.6794872 to the left, improve=0.01572268, (0 missing)
## CP < 0.3328947 to the left, improve=0.01453018, (0 missing)
## Surrogate splits:
## FCDi < 0.4228551 to the right, agree=0.660, adj=0.262, (0 split)
## CC < 0.4529334 to the right, agree=0.603, adj=0.138, (0 split)
## CP < 0.4369372 to the left, agree=0.603, adj=0.138, (0 split)
## FCDo < 0.269697 to the right, agree=0.582, adj=0.092, (0 split)
##
## Node number 5: 7 observations
## mean=1.103177, MSE=1.945491
##
## Node number 6: 29 observations
## mean=0.3183031, MSE=0.1521703
##
## Node number 7: 172 observations, complexity param=0.02239911
## mean=0.7992478, MSE=0.5076644
## left son=14 (26 obs) right son=15 (146 obs)
## Primary splits:
## CC < 0.6693845 to the right, improve=0.03664811, (0 missing)
## as.factor(CONCEITO) splits as LLLRR, improve=0.03207323, (0 missing)
## FCDi < 0.189441 to the right, improve=0.02366747, (0 missing)
## as.factor(DEPENDENCIAADM) splits as LR, improve=0.02292318, (0 missing)
## CP < 0.7083333 to the right, improve=0.01980541, (0 missing)
##
## Node number 8: 76 observations
## mean=0.3573047, MSE=0.1272899
##
## Node number 9: 65 observations, complexity param=0.01029784
## mean=0.586314, MSE=0.2782576
## left son=18 (55 obs) right son=19 (10 obs)
## Primary splits:
## FCDo < 0.6427432 to the left, improve=0.06189628, (0 missing)
## FCDi < 0.3582888 to the left, improve=0.05966784, (0 missing)
## CC < 0.6019898 to the right, improve=0.04629516, (0 missing)
## CP < 0.4375926 to the right, improve=0.02795421, (0 missing)
## as.factor(CONCEITO) splits as -L-RR, improve=0.01444040, (0 missing)
## Surrogate splits:
## FCDi < 0.7519201 to the left, agree=0.862, adj=0.1, (0 split)
##
## Node number 14: 26 observations
## mean=0.4760236, MSE=0.213218
##
## Node number 15: 146 observations, complexity param=0.02239911
## mean=0.8568083, MSE=0.5381819
## left son=30 (121 obs) right son=31 (25 obs)
## Primary splits:
## as.factor(CONCEITO) splits as LLLRR, improve=0.03076330, (0 missing)
## CC < 0.350456 to the right, improve=0.02644304, (0 missing)
## FCDi < 0.189441 to the right, improve=0.01612835, (0 missing)
## as.factor(DEPENDENCIAADM) splits as LR, improve=0.01338604, (0 missing)
## FCDo < 0.5465368 to the left, improve=0.01098118, (0 missing)
## Surrogate splits:
## CC < 0.3445861 to the right, agree=0.842, adj=0.08, (0 split)
## FCDo < 0.2344055 to the right, agree=0.836, adj=0.04, (0 split)
##
## Node number 18: 55 observations, complexity param=0.01029784
## mean=0.5303544, MSE=0.2043189
## left son=36 (14 obs) right son=37 (41 obs)
## Primary splits:
## CC < 0.6019898 to the right, improve=0.13981370, (0 missing)
## as.factor(CONCEITO) splits as -L-RR, improve=0.06595971, (0 missing)
## CP < 0.3640523 to the left, improve=0.06510467, (0 missing)
## FCDi < 0.3360526 to the left, improve=0.02538762, (0 missing)
## FCDo < 0.5857843 to the right, improve=0.02263407, (0 missing)
## Surrogate splits:
## FCDi < 0.6546053 to the right, agree=0.800, adj=0.214, (0 split)
## FCDo < 0.5522876 to the right, agree=0.782, adj=0.143, (0 split)
## as.factor(DEPENDENCIAADM) splits as LR, agree=0.782, adj=0.143, (0 split)
## CP < 0.3360417 to the left, agree=0.764, adj=0.071, (0 split)
##
## Node number 19: 10 observations
## mean=0.8940917, MSE=0.5729703
##
## Node number 30: 121 observations
## mean=0.7983214, MSE=0.3348285
##
## Node number 31: 25 observations, complexity param=0.02239911
## mean=1.139885, MSE=1.425724
## left son=62 (18 obs) right son=63 (7 obs)
## Primary splits:
## FCDi < 0.3356331 to the left, improve=1.217970e-01, (0 missing)
## FCDo < 0.5105856 to the left, improve=8.984864e-02, (0 missing)
## CP < 0.4816682 to the right, improve=3.685829e-02, (0 missing)
## CC < 0.4971226 to the right, improve=2.548008e-02, (0 missing)
## as.factor(CONCEITO) splits as ---RL, improve=3.728631e-05, (0 missing)
## Surrogate splits:
## FCDo < 0.5022523 to the left, agree=0.76, adj=0.143, (0 split)
##
## Node number 36: 14 observations
## mean=0.2411153, MSE=0.06078167
##
## Node number 37: 41 observations, complexity param=0.01029784
## mean=0.629119, MSE=0.2150106
## left son=74 (28 obs) right son=75 (13 obs)
## Primary splits:
## FCDi < 0.336875 to the left, improve=0.21413400, (0 missing)
## CC < 0.5347393 to the left, improve=0.11238680, (0 missing)
## FCDo < 0.5331439 to the left, improve=0.08401116, (0 missing)
## CP < 0.4375926 to the right, improve=0.06392309, (0 missing)
## as.factor(CONCEITO) splits as -L-RR, improve=0.05506055, (0 missing)
## Surrogate splits:
## CC < 0.5197655 to the left, agree=0.829, adj=0.462, (0 split)
## FCDo < 0.4796296 to the left, agree=0.805, adj=0.385, (0 split)
## CP < 0.342445 to the right, agree=0.707, adj=0.077, (0 split)
##
## Node number 62: 18 observations
## mean=0.8800191, MSE=0.4187633
##
## Node number 63: 7 observations
## mean=1.808111, MSE=3.394877
##
## Node number 74: 28 observations
## mean=0.4829129, MSE=0.1068154
##
## Node number 75: 13 observations
## mean=0.9440244, MSE=0.3028397
1.6 Produtos
fitProduto<- lm(formula = sqrt(SPPPr/QPPP) ~ FCDo + FCDi + CC + CP + as.factor(DEPENDENCIAADM) + as.factor(CONCEITO) + as.factor(UFPROGRAMA) + FCDi * CC - 1, data = no_outliers)
summary(fitProduto)##
## Call:
## lm(formula = sqrt(SPPPr/QPPP) ~ FCDo + FCDi + CC + CP + as.factor(DEPENDENCIAADM) +
## as.factor(CONCEITO) + as.factor(UFPROGRAMA) + FCDi * CC -
## 1, data = no_outliers)
##
## Residuals:
## Min 1Q Median 3Q Max
## -0.55609 -0.24614 -0.03405 0.15980 1.50068
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## FCDo 0.115563 0.130668 0.884 0.3772
## FCDi -0.222547 0.300259 -0.741 0.4591
## CC -0.250075 0.262536 -0.953 0.3416
## CP 0.008614 0.195205 0.044 0.9648
## as.factor(DEPENDENCIAADM)PRIVADA 0.436698 0.220394 1.981 0.0484 *
## as.factor(DEPENDENCIAADM)PÚBLICA 0.394790 0.209951 1.880 0.0610 .
## as.factor(CONCEITO)4 -0.014193 0.049457 -0.287 0.7743
## as.factor(CONCEITO)5 -0.084674 0.062648 -1.352 0.1775
## as.factor(CONCEITO)6 -0.162837 0.076814 -2.120 0.0348 *
## as.factor(CONCEITO)7 -0.204283 0.109380 -1.868 0.0627 .
## as.factor(UFPROGRAMA)AM 0.234963 0.230324 1.020 0.3084
## as.factor(UFPROGRAMA)AP -0.008091 0.355751 -0.023 0.9819
## as.factor(UFPROGRAMA)BA 0.026207 0.149458 0.175 0.8609
## as.factor(UFPROGRAMA)CE 0.020598 0.166463 0.124 0.9016
## as.factor(UFPROGRAMA)DF 0.027819 0.155581 0.179 0.8582
## as.factor(UFPROGRAMA)ES 0.047873 0.208212 0.230 0.8183
## as.factor(UFPROGRAMA)GO -0.057580 0.179537 -0.321 0.7486
## as.factor(UFPROGRAMA)MA -0.158047 0.209187 -0.756 0.4505
## as.factor(UFPROGRAMA)MG 0.100455 0.137588 0.730 0.4659
## as.factor(UFPROGRAMA)MS -0.150784 0.236881 -0.637 0.5249
## as.factor(UFPROGRAMA)MT -0.258484 0.270938 -0.954 0.3408
## as.factor(UFPROGRAMA)PA 0.068426 0.195390 0.350 0.7264
## as.factor(UFPROGRAMA)PB 0.222162 0.209568 1.060 0.2899
## as.factor(UFPROGRAMA)PE -0.022290 0.147913 -0.151 0.8803
## as.factor(UFPROGRAMA)PI -0.108230 0.229204 -0.472 0.6371
## as.factor(UFPROGRAMA)PR 0.095577 0.139740 0.684 0.4945
## as.factor(UFPROGRAMA)RJ 0.042602 0.141453 0.301 0.7635
## as.factor(UFPROGRAMA)RN -0.140447 0.152159 -0.923 0.3567
## as.factor(UFPROGRAMA)RO 0.173559 0.355771 0.488 0.6260
## as.factor(UFPROGRAMA)RS 0.257228 0.139063 1.850 0.0653 .
## as.factor(UFPROGRAMA)SC 0.158146 0.156568 1.010 0.3132
## as.factor(UFPROGRAMA)SE 0.048546 0.181207 0.268 0.7890
## as.factor(UFPROGRAMA)SP 0.124219 0.142141 0.874 0.3828
## as.factor(UFPROGRAMA)TO -0.056985 0.267566 -0.213 0.8315
## FCDi:CC 0.219510 0.477159 0.460 0.6458
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.331 on 314 degrees of freedom
## Multiple R-squared: 0.5408, Adjusted R-squared: 0.4896
## F-statistic: 10.56 on 35 and 314 DF, p-value: < 2.2e-161.6.0.1 Regressão de Poisson
regpoisson=glm(sqrt(SPPPr/QPPP) ~ FCDo + FCDi + CC + CP + as.factor(DEPENDENCIAADM) + as.factor(CONCEITO) + as.factor(UFPROGRAMA) + FCDi * CC -1, family="poisson", data= IMI_e_IPT)
summary(regpoisson)##
## Call:
## glm(formula = sqrt(SPPPr/QPPP) ~ FCDo + FCDi + CC + CP + as.factor(DEPENDENCIAADM) +
## as.factor(CONCEITO) + as.factor(UFPROGRAMA) + FCDi * CC -
## 1, family = "poisson", data = IMI_e_IPT)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -0.7312 -0.4113 -0.3454 -0.2339 2.6836
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## FCDo 1.3369 0.4763 2.807 0.005 **
## FCDi -0.8160 1.2430 -0.656 0.512
## CC -1.9591 1.0849 -1.806 0.071 .
## CP 0.3492 0.6544 0.534 0.594
## as.factor(DEPENDENCIAADM)PRIVADA -14.6500 726.8327 -0.020 0.984
## as.factor(DEPENDENCIAADM)PÚBLICA -15.0541 726.8327 -0.021 0.983
## as.factor(CONCEITO)4 0.1528 0.1781 0.858 0.391
## as.factor(CONCEITO)5 0.3060 0.2297 1.332 0.183
## as.factor(CONCEITO)6 0.1166 0.3205 0.364 0.716
## as.factor(CONCEITO)7 0.3040 0.4142 0.734 0.463
## as.factor(UFPROGRAMA)AL 12.5308 726.8327 0.017 0.986
## as.factor(UFPROGRAMA)AM 12.6555 726.8326 0.017 0.986
## as.factor(UFPROGRAMA)AP 12.6343 726.8350 0.017 0.986
## as.factor(UFPROGRAMA)BA 12.4648 726.8324 0.017 0.986
## as.factor(UFPROGRAMA)CE 12.5366 726.8324 0.017 0.986
## as.factor(UFPROGRAMA)DF 12.6250 726.8324 0.017 0.986
## as.factor(UFPROGRAMA)ES 11.8993 726.8327 0.016 0.987
## as.factor(UFPROGRAMA)GO 12.0363 726.8325 0.017 0.987
## as.factor(UFPROGRAMA)MA 11.5489 726.8329 0.016 0.987
## as.factor(UFPROGRAMA)MG 12.6873 726.8324 0.017 0.986
## as.factor(UFPROGRAMA)MS 11.7967 726.8329 0.016 0.987
## as.factor(UFPROGRAMA)MT 10.1730 726.8346 0.014 0.989
## as.factor(UFPROGRAMA)PA 12.8953 726.8324 0.018 0.986
## as.factor(UFPROGRAMA)PB 12.6954 726.8326 0.017 0.986
## as.factor(UFPROGRAMA)PE 12.0827 726.8325 0.017 0.987
## as.factor(UFPROGRAMA)PI 12.2559 726.8338 0.017 0.987
## as.factor(UFPROGRAMA)PR 12.8170 726.8324 0.018 0.986
## as.factor(UFPROGRAMA)RJ 12.5721 726.8324 0.017 0.986
## as.factor(UFPROGRAMA)RN 12.2530 726.8325 0.017 0.987
## as.factor(UFPROGRAMA)RO 13.6260 726.8327 0.019 0.985
## as.factor(UFPROGRAMA)RR -0.2529 1468.2756 0.000 1.000
## as.factor(UFPROGRAMA)RS 12.8192 726.8324 0.018 0.986
## as.factor(UFPROGRAMA)SC 12.7093 726.8324 0.017 0.986
## as.factor(UFPROGRAMA)SE 13.3714 726.8326 0.018 0.985
## as.factor(UFPROGRAMA)SP 12.3229 726.8324 0.017 0.986
## as.factor(UFPROGRAMA)TO 12.3185 726.8329 0.017 0.986
## FCDi:CC 1.4643 2.0123 0.728 0.467
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for poisson family taken to be 1)
##
## Null deviance: 4910.85 on 2858 degrees of freedom
## Residual deviance: 709.79 on 2821 degrees of freedom
## AIC: Inf
##
## Number of Fisher Scoring iterations: 131.6.0.2 Árvore de decisão
arvore_produto <- rpart(SPPPr/QPPP ~ FCDo + FCDi + CC + CP + as.factor(DEPENDENCIAADM) + as.factor(CONCEITO), data = no_outliers)
summary(arvore_produto)## Call:
## rpart(formula = SPPPr/QPPP ~ FCDo + FCDi + CC + CP + as.factor(DEPENDENCIAADM) +
## as.factor(CONCEITO), data = no_outliers)
## n= 349
##
## CP nsplit rel error xerror xstd
## 1 0.03178604 0 1.0000000 1.008583 0.3188953
## 2 0.01939137 1 0.9682140 1.105910 0.3215916
## 3 0.01489306 3 0.9294312 1.180256 0.3190358
## 4 0.01000000 6 0.8847520 1.240110 0.3232844
##
## Variable importance
## FCDi CP CC as.factor(CONCEITO)
## 40 31 14 8
## FCDo
## 6
##
## Node number 1: 349 observations, complexity param=0.03178604
## mean=0.214688, MSE=0.1752302
## left son=2 (339 obs) right son=3 (10 obs)
## Primary splits:
## FCDi < 0.9615385 to the left, improve=0.03178604, (0 missing)
## as.factor(DEPENDENCIAADM) splits as RL, improve=0.02574168, (0 missing)
## CP < 0.4987637 to the left, improve=0.01765411, (0 missing)
## FCDo < 0.4122316 to the left, improve=0.01376318, (0 missing)
## as.factor(CONCEITO) splits as RRLLL, improve=0.01106026, (0 missing)
##
## Node number 2: 339 observations, complexity param=0.01939137
## mean=0.2018699, MSE=0.1308178
## left son=4 (307 obs) right son=5 (32 obs)
## Primary splits:
## CP < 0.550463 to the left, improve=0.014634150, (0 missing)
## FCDo < 0.4122316 to the left, improve=0.013348420, (0 missing)
## as.factor(DEPENDENCIAADM) splits as RL, improve=0.011491610, (0 missing)
## as.factor(CONCEITO) splits as RRRLL, improve=0.010421320, (0 missing)
## FCDi < 0.3702381 to the left, improve=0.009293826, (0 missing)
##
## Node number 3: 10 observations
## mean=0.6492208, MSE=1.486423
##
## Node number 4: 307 observations, complexity param=0.01489306
## mean=0.1877438, MSE=0.09350723
## left son=8 (204 obs) right son=9 (103 obs)
## Primary splits:
## as.factor(CONCEITO) splits as RLLLL, improve=0.017920130, (0 missing)
## FCDo < 0.8257576 to the left, improve=0.014202210, (0 missing)
## FCDi < 0.4205263 to the left, improve=0.011737620, (0 missing)
## CP < 0.5413492 to the right, improve=0.008810057, (0 missing)
## CC < 0.650146 to the left, improve=0.006771424, (0 missing)
## Surrogate splits:
## FCDi < 0.5732143 to the left, agree=0.785, adj=0.359, (0 split)
## FCDo < 0.6376263 to the left, agree=0.707, adj=0.126, (0 split)
## CP < 0.3026042 to the right, agree=0.678, adj=0.039, (0 split)
##
## Node number 5: 32 observations, complexity param=0.01939137
## mean=0.3373924, MSE=0.4684851
## left son=10 (24 obs) right son=11 (8 obs)
## Primary splits:
## CP < 0.5688095 to the right, improve=0.11491750, (0 missing)
## as.factor(DEPENDENCIAADM) splits as RL, improve=0.09985498, (0 missing)
## as.factor(CONCEITO) splits as LRLL-, improve=0.09578208, (0 missing)
## FCDi < 0.5811404 to the right, improve=0.04275088, (0 missing)
## CC < 0.5961445 to the right, improve=0.04187705, (0 missing)
## Surrogate splits:
## as.factor(CONCEITO) splits as LLRL-, agree=0.781, adj=0.125, (0 split)
##
## Node number 8: 204 observations
## mean=0.1586569, MSE=0.05966425
##
## Node number 9: 103 observations, complexity param=0.01489306
## mean=0.2453527, MSE=0.1555416
## left son=18 (78 obs) right son=19 (25 obs)
## Primary splits:
## CC < 0.5289366 to the right, improve=0.0571028000, (0 missing)
## FCDi < 0.5732143 to the right, improve=0.0382717800, (0 missing)
## FCDo < 0.6171498 to the right, improve=0.0280340700, (0 missing)
## CP < 0.5246181 to the right, improve=0.0181320300, (0 missing)
## as.factor(DEPENDENCIAADM) splits as RL, improve=0.0006905795, (0 missing)
## Surrogate splits:
## FCDi < 0.251179 to the right, agree=0.777, adj=0.08, (0 split)
## FCDo < 0.3444444 to the right, agree=0.767, adj=0.04, (0 split)
##
## Node number 10: 24 observations
## mean=0.2034308, MSE=0.1841712
##
## Node number 11: 8 observations
## mean=0.7392771, MSE=1.106078
##
## Node number 18: 78 observations
## mean=0.1919977, MSE=0.1060473
##
## Node number 19: 25 observations, complexity param=0.01489306
## mean=0.4118201, MSE=0.2733707
## left son=38 (12 obs) right son=39 (13 obs)
## Primary splits:
## FCDi < 0.5668241 to the right, improve=0.19067320, (0 missing)
## FCDo < 0.5358974 to the left, improve=0.10093880, (0 missing)
## CC < 0.5011218 to the left, improve=0.09637133, (0 missing)
## CP < 0.479386 to the left, improve=0.03544710, (0 missing)
## Surrogate splits:
## FCDo < 0.5694444 to the right, agree=0.68, adj=0.333, (0 split)
## CC < 0.5011218 to the left, agree=0.64, adj=0.250, (0 split)
## CP < 0.4449653 to the right, agree=0.64, adj=0.250, (0 split)
##
## Node number 38: 12 observations
## mean=0.1741898, MSE=0.03724166
##
## Node number 39: 13 observations
## mean=0.6311712, MSE=0.3910966
1.7 Aplicativos
fitApp<- lm(formula = sqrt(SAPP / QPPP) ~ FCDo + FCDi + CC + CP + as.factor(DEPENDENCIAADM) + as.factor(CONCEITO) + as.factor(UFPROGRAMA) + FCDi * CC - 1, data = no_outliers)
summary(fitApp)##
## Call:
## lm(formula = sqrt(SAPP/QPPP) ~ FCDo + FCDi + CC + CP + as.factor(DEPENDENCIAADM) +
## as.factor(CONCEITO) + as.factor(UFPROGRAMA) + FCDi * CC -
## 1, data = no_outliers)
##
## Residuals:
## Min 1Q Median 3Q Max
## -0.70947 -0.26896 -0.08895 0.24785 1.38975
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## FCDo -0.331245 0.152323 -2.175 0.0304 *
## FCDi 0.217738 0.350020 0.622 0.5343
## CC 0.750816 0.306045 2.453 0.0147 *
## CP 0.052085 0.227556 0.229 0.8191
## as.factor(DEPENDENCIAADM)PRIVADA 0.096309 0.256919 0.375 0.7080
## as.factor(DEPENDENCIAADM)PÚBLICA 0.011137 0.244746 0.046 0.9637
## as.factor(CONCEITO)4 -0.026093 0.057654 -0.453 0.6512
## as.factor(CONCEITO)5 -0.130041 0.073031 -1.781 0.0759 .
## as.factor(CONCEITO)6 -0.130877 0.089544 -1.462 0.1449
## as.factor(CONCEITO)7 -0.302464 0.127508 -2.372 0.0183 *
## as.factor(UFPROGRAMA)AM 0.360370 0.268495 1.342 0.1805
## as.factor(UFPROGRAMA)AP -0.265171 0.414709 -0.639 0.5230
## as.factor(UFPROGRAMA)BA 0.004714 0.174227 0.027 0.9784
## as.factor(UFPROGRAMA)CE 0.132524 0.194050 0.683 0.4952
## as.factor(UFPROGRAMA)DF 0.069679 0.181366 0.384 0.7011
## as.factor(UFPROGRAMA)ES 0.131945 0.242719 0.544 0.5871
## as.factor(UFPROGRAMA)GO -0.012691 0.209291 -0.061 0.9517
## as.factor(UFPROGRAMA)MA 0.323108 0.243855 1.325 0.1861
## as.factor(UFPROGRAMA)MG 0.141139 0.160390 0.880 0.3795
## as.factor(UFPROGRAMA)MS 0.091445 0.276138 0.331 0.7407
## as.factor(UFPROGRAMA)MT -0.224905 0.315840 -0.712 0.4769
## as.factor(UFPROGRAMA)PA -0.091803 0.227771 -0.403 0.6872
## as.factor(UFPROGRAMA)PB -0.124886 0.244299 -0.511 0.6096
## as.factor(UFPROGRAMA)PE -0.005428 0.172426 -0.031 0.9749
## as.factor(UFPROGRAMA)PI -0.030504 0.267189 -0.114 0.9092
## as.factor(UFPROGRAMA)PR 0.081857 0.162899 0.502 0.6157
## as.factor(UFPROGRAMA)RJ 0.373183 0.164896 2.263 0.0243 *
## as.factor(UFPROGRAMA)RN 0.031056 0.177376 0.175 0.8611
## as.factor(UFPROGRAMA)RO 0.038910 0.414732 0.094 0.9253
## as.factor(UFPROGRAMA)RS 0.150616 0.162110 0.929 0.3536
## as.factor(UFPROGRAMA)SC 0.205903 0.182516 1.128 0.2601
## as.factor(UFPROGRAMA)SE 0.022686 0.211238 0.107 0.9145
## as.factor(UFPROGRAMA)SP 0.110609 0.165697 0.668 0.5049
## as.factor(UFPROGRAMA)TO 0.445293 0.311909 1.428 0.1544
## FCDi:CC -0.387726 0.556237 -0.697 0.4863
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.3859 on 314 degrees of freedom
## Multiple R-squared: 0.5485, Adjusted R-squared: 0.4982
## F-statistic: 10.9 on 35 and 314 DF, p-value: < 2.2e-161.7.0.1 Regressão de Poisson
regpoisson=glm(sqrt(SAPP/QPPP) ~ FCDo + FCDi + CC + CP + as.factor(DEPENDENCIAADM) + as.factor(CONCEITO) + as.factor(UFPROGRAMA) + FCDi * CC -1, family="poisson", data= IMI_e_IPT)
summary(regpoisson)##
## Call:
## glm(formula = sqrt(SAPP/QPPP) ~ FCDo + FCDi + CC + CP + as.factor(DEPENDENCIAADM) +
## as.factor(CONCEITO) + as.factor(UFPROGRAMA) + FCDi * CC -
## 1, family = "poisson", data = IMI_e_IPT)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -0.8264 -0.4847 -0.4213 0.1997 2.6427
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## FCDo 8.209e-01 3.857e-01 2.128 0.0333 *
## FCDi 1.134e+00 9.870e-01 1.149 0.2505
## CC 5.941e-01 8.224e-01 0.722 0.4701
## CP -8.244e-01 6.099e-01 -1.352 0.1765
## as.factor(DEPENDENCIAADM)PRIVADA -1.570e+01 7.290e+02 -0.022 0.9828
## as.factor(DEPENDENCIAADM)PÚBLICA -1.598e+01 7.290e+02 -0.022 0.9825
## as.factor(CONCEITO)4 1.992e-01 1.436e-01 1.387 0.1654
## as.factor(CONCEITO)5 8.424e-02 2.023e-01 0.416 0.6771
## as.factor(CONCEITO)6 3.620e-01 2.511e-01 1.441 0.1494
## as.factor(CONCEITO)7 8.261e-02 3.805e-01 0.217 0.8281
## as.factor(UFPROGRAMA)AL 1.319e+01 7.290e+02 0.018 0.9856
## as.factor(UFPROGRAMA)AM 1.251e+01 7.290e+02 0.017 0.9863
## as.factor(UFPROGRAMA)AP -2.363e-02 1.036e+03 0.000 1.0000
## as.factor(UFPROGRAMA)BA 1.302e+01 7.290e+02 0.018 0.9858
## as.factor(UFPROGRAMA)CE 1.297e+01 7.290e+02 0.018 0.9858
## as.factor(UFPROGRAMA)DF 1.319e+01 7.290e+02 0.018 0.9856
## as.factor(UFPROGRAMA)ES 1.280e+01 7.290e+02 0.018 0.9860
## as.factor(UFPROGRAMA)GO 1.281e+01 7.290e+02 0.018 0.9860
## as.factor(UFPROGRAMA)MA 1.286e+01 7.290e+02 0.018 0.9859
## as.factor(UFPROGRAMA)MG 1.321e+01 7.290e+02 0.018 0.9855
## as.factor(UFPROGRAMA)MS 1.256e+01 7.290e+02 0.017 0.9863
## as.factor(UFPROGRAMA)MT 1.223e+01 7.290e+02 0.017 0.9866
## as.factor(UFPROGRAMA)PA 1.277e+01 7.290e+02 0.018 0.9860
## as.factor(UFPROGRAMA)PB 1.232e+01 7.290e+02 0.017 0.9865
## as.factor(UFPROGRAMA)PE 1.293e+01 7.290e+02 0.018 0.9858
## as.factor(UFPROGRAMA)PI 1.305e+01 7.290e+02 0.018 0.9857
## as.factor(UFPROGRAMA)PR 1.311e+01 7.290e+02 0.018 0.9856
## as.factor(UFPROGRAMA)RJ 1.324e+01 7.290e+02 0.018 0.9855
## as.factor(UFPROGRAMA)RN 1.323e+01 7.290e+02 0.018 0.9855
## as.factor(UFPROGRAMA)RO 1.163e+01 7.290e+02 0.016 0.9873
## as.factor(UFPROGRAMA)RR 1.688e-03 1.469e+03 0.000 1.0000
## as.factor(UFPROGRAMA)RS 1.322e+01 7.290e+02 0.018 0.9855
## as.factor(UFPROGRAMA)SC 1.320e+01 7.290e+02 0.018 0.9856
## as.factor(UFPROGRAMA)SE 1.349e+01 7.290e+02 0.019 0.9852
## as.factor(UFPROGRAMA)SP 1.293e+01 7.290e+02 0.018 0.9858
## as.factor(UFPROGRAMA)TO 1.351e+01 7.290e+02 0.019 0.9852
## FCDi:CC -1.298e+00 1.548e+00 -0.839 0.4017
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for poisson family taken to be 1)
##
## Null deviance: 4605.88 on 2858 degrees of freedom
## Residual deviance: 886.23 on 2821 degrees of freedom
## AIC: Inf
##
## Number of Fisher Scoring iterations: 131.7.0.2 Árvore de decisão
arvore_aplicativo <- rpart(SAPP/QPPP ~ FCDo + FCDi + CC + CP + as.factor(DEPENDENCIAADM) + as.factor(CONCEITO), data = no_outliers)
summary(arvore_aplicativo)## Call:
## rpart(formula = SAPP/QPPP ~ FCDo + FCDi + CC + CP + as.factor(DEPENDENCIAADM) +
## as.factor(CONCEITO), data = no_outliers)
## n= 349
##
## CP nsplit rel error xerror xstd
## 1 0.06860452 0 1.0000000 1.005998 0.1579283
## 2 0.04033968 1 0.9313955 1.072299 0.1546801
## 3 0.02909134 2 0.8910558 1.050254 0.1533450
## 4 0.02413469 4 0.8328731 1.094798 0.1607488
## 5 0.02033243 5 0.8087384 1.111962 0.1607320
## 6 0.01369760 7 0.7680736 1.132793 0.1569157
## 7 0.01356819 9 0.7406784 1.116188 0.1527953
## 8 0.01055298 12 0.6999738 1.133826 0.1538961
## 9 0.01000000 13 0.6894208 1.144857 0.1546434
##
## Variable importance
## CC CP FCDo
## 29 29 18
## as.factor(CONCEITO) FCDi as.factor(DEPENDENCIAADM)
## 12 11 2
##
## Node number 1: 349 observations, complexity param=0.06860452
## mean=0.2967639, MSE=0.2019442
## left son=2 (229 obs) right son=3 (120 obs)
## Primary splits:
## CC < 0.6032122 to the left, improve=0.06860452, (0 missing)
## CP < 0.78125 to the left, improve=0.06359468, (0 missing)
## FCDi < 0.4027376 to the left, improve=0.03325919, (0 missing)
## as.factor(DEPENDENCIAADM) splits as RL, improve=0.02282198, (0 missing)
## FCDo < 0.739418 to the left, improve=0.02128091, (0 missing)
## Surrogate splits:
## FCDo < 0.6933761 to the left, agree=0.728, adj=0.208, (0 split)
## FCDi < 0.6287683 to the left, agree=0.688, adj=0.092, (0 split)
## as.factor(DEPENDENCIAADM) splits as RL, agree=0.676, adj=0.058, (0 split)
## CP < 0.78125 to the left, agree=0.665, adj=0.025, (0 split)
##
## Node number 2: 229 observations, complexity param=0.02413469
## mean=0.211559, MSE=0.118884
## left son=4 (166 obs) right son=5 (63 obs)
## Primary splits:
## CP < 0.4212456 to the right, improve=0.06247979, (0 missing)
## as.factor(CONCEITO) splits as RRLLL, improve=0.04016652, (0 missing)
## CC < 0.4374782 to the left, improve=0.03282737, (0 missing)
## FCDi < 0.2736185 to the left, improve=0.02454798, (0 missing)
## FCDo < 0.739418 to the left, improve=0.01621318, (0 missing)
## Surrogate splits:
## FCDo < 0.2384259 to the right, agree=0.747, adj=0.079, (0 split)
## FCDi < 1.046875 to the left, agree=0.729, adj=0.016, (0 split)
##
## Node number 3: 120 observations, complexity param=0.04033968
## mean=0.4593634, MSE=0.3201578
## left son=6 (107 obs) right son=7 (13 obs)
## Primary splits:
## CP < 0.5533399 to the left, improve=0.074002070, (0 missing)
## FCDi < 0.729021 to the right, improve=0.036014390, (0 missing)
## CC < 0.6384672 to the right, improve=0.025727530, (0 missing)
## FCDo < 0.6399573 to the right, improve=0.025400750, (0 missing)
## as.factor(DEPENDENCIAADM) splits as RL, improve=0.008366206, (0 missing)
##
## Node number 4: 166 observations, complexity param=0.01055298
## mean=0.1584647, MSE=0.08188266
## left son=8 (154 obs) right son=9 (12 obs)
## Primary splits:
## FCDo < 0.739418 to the left, improve=0.05471820, (0 missing)
## as.factor(CONCEITO) splits as RRLLL, improve=0.05190366, (0 missing)
## FCDi < 0.3964286 to the left, improve=0.03873325, (0 missing)
## as.factor(DEPENDENCIAADM) splits as RL, improve=0.03476798, (0 missing)
## CP < 0.4971769 to the right, improve=0.03366370, (0 missing)
##
## Node number 5: 63 observations, complexity param=0.02033243
## mean=0.3514582, MSE=0.1893799
## left son=10 (42 obs) right son=11 (21 obs)
## Primary splits:
## CC < 0.5442195 to the left, improve=0.09792120, (0 missing)
## FCDi < 0.2716161 to the left, improve=0.08624086, (0 missing)
## as.factor(CONCEITO) splits as RRRLL, improve=0.03867725, (0 missing)
## CP < 0.3452536 to the left, improve=0.01296667, (0 missing)
## FCDo < 0.6386946 to the right, improve=0.01108625, (0 missing)
##
## Node number 6: 107 observations, complexity param=0.02909134
## mean=0.4057117, MSE=0.2309145
## left son=12 (82 obs) right son=13 (25 obs)
## Primary splits:
## CP < 0.4050694 to the right, improve=0.08159447, (0 missing)
## CC < 0.6247379 to the right, improve=0.03772789, (0 missing)
## as.factor(CONCEITO) splits as LLRRL, improve=0.03368993, (0 missing)
## FCDi < 0.6636905 to the right, improve=0.03148115, (0 missing)
## FCDo < 0.6125541 to the right, improve=0.01568586, (0 missing)
## Surrogate splits:
## FCDo < 0.925 to the left, agree=0.785, adj=0.08, (0 split)
##
## Node number 7: 13 observations
## mean=0.9009584, MSE=0.8360003
##
## Node number 8: 154 observations
## mean=0.1397797, MSE=0.07479888
##
## Node number 9: 12 observations
## mean=0.398255, MSE=0.1108114
##
## Node number 10: 42 observations
## mean=0.2551661, MSE=0.09444689
##
## Node number 11: 21 observations, complexity param=0.02033243
## mean=0.5440422, MSE=0.323613
## left son=22 (11 obs) right son=23 (10 obs)
## Primary splits:
## CC < 0.5671054 to the right, improve=0.24981450, (0 missing)
## FCDi < 0.4063129 to the right, improve=0.14230470, (0 missing)
## CP < 0.3773148 to the right, improve=0.09521619, (0 missing)
## FCDo < 0.4768519 to the right, improve=0.05262435, (0 missing)
## Surrogate splits:
## CP < 0.3773148 to the right, agree=0.714, adj=0.4, (0 split)
## FCDo < 0.4768519 to the right, agree=0.667, adj=0.3, (0 split)
## as.factor(CONCEITO) splits as RLLL-, agree=0.667, adj=0.3, (0 split)
## FCDi < 0.3301574 to the right, agree=0.619, adj=0.2, (0 split)
## as.factor(DEPENDENCIAADM) splits as RL, agree=0.571, adj=0.1, (0 split)
##
## Node number 12: 82 observations, complexity param=0.0136976
## mean=0.3299204, MSE=0.1778742
## left son=24 (59 obs) right son=25 (23 obs)
## Primary splits:
## FCDo < 0.737037 to the left, improve=0.06593106, (0 missing)
## CP < 0.4119817 to the left, improve=0.05386413, (0 missing)
## as.factor(DEPENDENCIAADM) splits as RL, improve=0.02959628, (0 missing)
## FCDi < 0.7236364 to the right, improve=0.02461049, (0 missing)
## CC < 0.6184748 to the right, improve=0.01630720, (0 missing)
## Surrogate splits:
## CP < 0.533637 to the left, agree=0.780, adj=0.217, (0 split)
## FCDi < 0.7703297 to the left, agree=0.744, adj=0.087, (0 split)
## CC < 0.6044909 to the right, agree=0.732, adj=0.043, (0 split)
##
## Node number 13: 25 observations, complexity param=0.02909134
## mean=0.6543069, MSE=0.3242456
## left son=26 (14 obs) right son=27 (11 obs)
## Primary splits:
## as.factor(CONCEITO) splits as LRRR-, improve=0.25716370, (0 missing)
## CC < 0.749 to the right, improve=0.18471290, (0 missing)
## FCDo < 0.5798319 to the right, improve=0.16045070, (0 missing)
## CP < 0.3686075 to the left, improve=0.06953281, (0 missing)
## FCDi < 0.5083333 to the right, improve=0.06574935, (0 missing)
## Surrogate splits:
## FCDo < 0.5634921 to the right, agree=0.76, adj=0.455, (0 split)
## FCDi < 0.5083333 to the right, agree=0.76, adj=0.455, (0 split)
## CP < 0.3657617 to the right, agree=0.64, adj=0.182, (0 split)
## CC < 0.6129171 to the right, agree=0.60, adj=0.091, (0 split)
##
## Node number 22: 11 observations
## mean=0.2729448, MSE=0.09709803
##
## Node number 23: 10 observations
## mean=0.8422494, MSE=0.4030088
##
## Node number 24: 59 observations, complexity param=0.01356819
## mean=0.262306, MSE=0.1268679
## left son=48 (16 obs) right son=49 (43 obs)
## Primary splits:
## FCDo < 0.6742424 to the right, improve=0.11466020, (0 missing)
## CC < 0.722433 to the left, improve=0.05258454, (0 missing)
## as.factor(CONCEITO) splits as RLLRL, improve=0.04397665, (0 missing)
## FCDi < 0.7236364 to the right, improve=0.04356117, (0 missing)
## CP < 0.4240385 to the right, improve=0.01656819, (0 missing)
##
## Node number 25: 23 observations, complexity param=0.0136976
## mean=0.5033662, MSE=0.2669057
## left son=50 (10 obs) right son=51 (13 obs)
## Primary splits:
## as.factor(CONCEITO) splits as LRR--, improve=0.15786770, (0 missing)
## FCDi < 0.5418752 to the right, improve=0.13674940, (0 missing)
## CP < 0.4524184 to the left, improve=0.10071030, (0 missing)
## FCDo < 0.781746 to the right, improve=0.06780701, (0 missing)
## as.factor(DEPENDENCIAADM) splits as RL, improve=0.02786219, (0 missing)
## Surrogate splits:
## FCDi < 0.5965812 to the right, agree=0.739, adj=0.4, (0 split)
## CP < 0.424625 to the left, agree=0.696, adj=0.3, (0 split)
## FCDo < 0.7541478 to the left, agree=0.652, adj=0.2, (0 split)
## CC < 0.639881 to the right, agree=0.609, adj=0.1, (0 split)
## as.factor(DEPENDENCIAADM) splits as RL, agree=0.609, adj=0.1, (0 split)
##
## Node number 26: 14 observations
## mean=0.3983457, MSE=0.130211
##
## Node number 27: 11 observations
## mean=0.9800758, MSE=0.3816892
##
## Node number 48: 16 observations
## mean=0.06458333, MSE=0.02259983
##
## Node number 49: 43 observations, complexity param=0.01356819
## mean=0.3358772, MSE=0.1457059
## left son=98 (16 obs) right son=99 (27 obs)
## Primary splits:
## CC < 0.6647727 to the left, improve=0.14452500, (0 missing)
## FCDo < 0.5962963 to the left, improve=0.11837030, (0 missing)
## as.factor(CONCEITO) splits as RLLRL, improve=0.09244643, (0 missing)
## FCDi < 0.7253846 to the right, improve=0.05993227, (0 missing)
## CP < 0.4830791 to the left, improve=0.04589235, (0 missing)
## Surrogate splits:
## as.factor(CONCEITO) splits as RRLRL, agree=0.721, adj=0.250, (0 split)
## FCDi < 0.3690172 to the left, agree=0.674, adj=0.125, (0 split)
## CP < 0.4094742 to the left, agree=0.651, adj=0.063, (0 split)
##
## Node number 50: 10 observations
## mean=0.2693223, MSE=0.09933087
##
## Node number 51: 13 observations
## mean=0.6833999, MSE=0.3212614
##
## Node number 98: 16 observations
## mean=0.1473683, MSE=0.0727452
##
## Node number 99: 27 observations, complexity param=0.01356819
## mean=0.4475862, MSE=0.1554048
## left son=198 (7 obs) right son=199 (20 obs)
## Primary splits:
## FCDi < 0.7253846 to the right, improve=0.26336040, (0 missing)
## FCDo < 0.5962963 to the left, improve=0.16101660, (0 missing)
## CC < 0.8583423 to the right, improve=0.08094265, (0 missing)
## as.factor(CONCEITO) splits as RLLR-, improve=0.05372785, (0 missing)
## as.factor(DEPENDENCIAADM) splits as LR, improve=0.02282374, (0 missing)
## Surrogate splits:
## CP < 0.4170718 to the left, agree=0.889, adj=0.571, (0 split)
##
## Node number 198: 7 observations
## mean=0.1056277, MSE=0.01153202
##
## Node number 199: 20 observations
## mean=0.5672717, MSE=0.1505082
1.8 Conclusões
As arvores de decisão paracem ser melhor para representar as relações entre os indicadores do modelo, pois não seguem a lógica linear e apresentam particularidades mais detalhadas.A regressão de Poisson provê um modelo mais ajustado.
2 Agradecimentos
Ao Instituto Stela, à UFSC, à CAPES e ao CNPq.
3 Referências
Checking normality for parametric tests in R https://www.sheffield.ac.uk/polopoly_fs/1.579191!/file/stcp-karadimitriou-normalR.pdf
Normality Test in R https://www.datanovia.com/en/lessons/normality-test-in-r/
Como realizar teste de normalidade no R ? https://rpubs.com/paternogbc/46768
Fazendo os testes de Kolmogorov-Smirnov e de Shapiro-Wilk para normalidade http://www.dpi.ufv.br/~peternelli/tutoriaisR/tutoriaisRempdf/tutorial.KS.SW.normalidade.11112004.pdf
BIOESTATÍSTICA USANDO R https://cran.r-project.org/doc/contrib/Beasley-BioestatisticaUsandoR.pdf
Delineamentos Experimentais https://smolski.github.io/livroavancado/analisf.html
Regression Models in R Multicollinearity in R https://datascienceplus.com/multicollinearity-in-r/
Multicollinearity in R https://www.rpubs.com/dudubiologico/545528
Ajuste de Modelos Não Lineares http://www.leg.ufpr.br/~walmes/cursoR/mgest/3reg-nao-linear.html
Tutorial — Ajuste e Interpretação de Regressão Linear com R https://medium.com/data-hackers/tutorial-ajuste-e-interpreta%C3%A7%C3%A3o-de-regress%C3%A3o-linear-com-r-5b23c4ddb72
CURSO - Modelos de regressão não linear https://www.ime.unicamp.br/~cnaber/cursomodelosnaolinearesR.pdf
Aplicação de modelos de regressão linear e não linear em ciências agrárias http://www.leg.ufpr.br/~walmes/cursoR/cnpaf3/cnpaf02trailer.html
Recursos Computacionais Utilizando R http://www.dex.ufla.br/~danielff/meusarquivospdf/RRC0.pdf
Modelos Não Lineares e suas Aplicações https://www.ufjf.br/cursoestatistica/files/2014/04/Modelos-N%c3%a3o-Lineares-e-suas-Aplica%c3%a7%c3%b5es.pdf
Modeloagem - Aprendizado Estatístico http://material.curso-r.com/modelos/
MODELOS DE REGRESSÃO- com apoio computacional https://www.ime.unicamp.br/~cnaber/Livro_MLG.pdf
MODELOS DE REGRESSÃO LINEARES PARA ESTIMATIVA DE PRODUTIVIDADE DA SOJA NO OESTE DO PARANÁ, UTILIZANDO DADOS ESPECTRAIS https://www.scielo.br/pdf/eagri/v30n3/14.pdf
Aplicação do Teste de Farrar-Glauber para Análise de Multicolinearidade Em Regressões Lineares https://ibape-nacional.com.br/biblioteca/wp-content/uploads/2020/02/AO-27-Aplica%C3%A7%C3%A3o-do-Teste-de-Farrar-Glauber-para-An%C3%A1lise.pdf
Regressão Logística: O método estatístico mais utilizado para modelar variáveis categóricas. https://matheusfacure.github.io/2017/02/25/regr-log/
Linear Regression http://rstudio-pubs-static.s3.amazonaws.com/428179_4d1959eb7bda4ed1b9ae5bb86004eae3.html
Regression http://www.mat.ufrgs.br/~giacomo/Softwares/R/Crawley/Crawley%20-%20The%20Book%20R/ch10.pdf
Regressão de Poisson https://smolski.github.io/livroavancado/regressao-de-poisson.html
Tutorial — Ajuste e Interpretação de Regressão Linear com R https://medium.com/data-hackers/tutorial-ajuste-e-interpreta%C3%A7%C3%A3o-de-regress%C3%A3o-linear-com-r-5b23c4ddb72
Estatística Prática para Docentes e Pós-Graduandos de Geraldo Maia Campos 11. Aditividade e homogeneidade http://www.forp.usp.br/restauradora/gmc/gmc_livro/gmc_livro_cap11.html
TESTES DE NORMALIDADE EM ANÁLISES ESTATÍSTICAS: UMA ORIENTAÇÃO PARA PRATICANTES EM CIÊNCIAS DA SAÚDE E ATIVIDADE FÍSICA file:///C:/Users/Jacob/Documents/R/6583-Texto%20do%20artigo-43438-1-10-20171008.pdf
regressão logística https://www.rpubs.com/dudubiologico/545528
Teste para normalidade e homocedasticidade https://biostatistics-uem.github.io/Bio/aula8/teste_normalidade_homocedasticidade.html#:~:text=Em%20an%C3%A1lise%20de%20vari%C3%A2ncia(ANOVA,que%20a%20ANOVA%20tenha%20validade.