El diseño muestral es el que se logró armar considerando los diseños de cada país Finalmente se trabaja con un diseño equivalente a 2 etapas, primera donde las psu son las escuelas, el pais es la variable de estratficacióny se usan como pesos muestrales los que se consideraron en Uruguay (ver diseño de Uruguay y para POA se creó un nuevo vector de peso usando como variables de ajuste distribucion por sexo y por tipo de escuela)
LOs analisis son unificados (no por pais)
names(disenio_urubra$variables) [1] “id” “pais” “ficha”
[4] “eta” “tipoesc” “sexoinv”
[7] “weight” “weight.rec” “numesc”
[10] “numer_esc” “idade” “socioecon4cat”
[13] “socioecon3cat” “escolmae” “escolmae13cat”
[16] “escolmaerecat23cat” “freqescov” “freqescovacat”
[19] “usofio” “freqfio” “usocreme”
[22] “idadcreme” “visidentcatonde” “visiquando”
[25] “fluorprof” “idadefluor” “isg.”
[28] “isg45” “isg20” “cposoms”
[31] “cpodoms” “prevoms” “cposbere”
[34] “cpodbere” “prevbere” “cposicdas”
[37] “cpodicdas” “previcdas”
load("~/Dropbox/odontologia/maestria_anunziatta/BancoUnidoBrasUru/BRASURU(11102013).RData")
library(survey)
## Attaching package: 'survey'
## The following object(s) are masked from 'package:graphics':
##
## dotchart
library(car)
## Loading required package: MASS
## Loading required package: nnet
# 11 de octubre 2013
modelo0.bin <- svyglm(prevoms ~ pais, disenio_urubra, family = quasibinomial())
summary(modelo0.bin)
##
## Call:
## svyglm(formula = prevoms ~ pais, disenio_urubra, family = quasibinomial())
##
## Survey design:
## svydesign(id = ~numer_esc, strata = ~pais, weights = ~round(weight.rec,
## 1), data = brasuru, nest = TRUE)
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 0.2185 0.0952 2.30 0.024 *
## paisUru 0.2484 0.1420 1.75 0.084 .
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for quasibinomial family taken to be 1)
##
## Number of Fisher Scoring iterations: 4
confint(modelo0.bin)
## 2.5 % 97.5 %
## (Intercept) 0.03195 0.4050
## paisUru -0.02999 0.5268
exp(modelo0.bin$coefficients)
## (Intercept) paisUru
## 1.244 1.282
modelo1.bin <- svyglm(prevoms ~ sexoinv, disenio_urubra, family = quasibinomial())
summary(modelo1.bin)
##
## Call:
## svyglm(formula = prevoms ~ sexoinv, disenio_urubra, family = quasibinomial())
##
## Survey design:
## svydesign(id = ~numer_esc, strata = ~pais, weights = ~round(weight.rec,
## 1), data = brasuru, nest = TRUE)
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 0.316 0.101 3.13 0.0025 **
## sexoinv2-F 0.087 0.127 0.68 0.4961
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for quasibinomial family taken to be 1)
##
## Number of Fisher Scoring iterations: 4
confint(modelo1.bin)
## 2.5 % 97.5 %
## (Intercept) 0.1180 0.5148
## sexoinv2-F -0.1623 0.3362
exp(modelo1.bin$coefficients)
## (Intercept) sexoinv2-F
## 1.372 1.091
modelo2.bin <- svyglm(prevoms ~ socioecon4cat, disenio_urubra, family = quasibinomial())
summary(modelo2.bin)
##
## Call:
## svyglm(formula = prevoms ~ socioecon4cat, disenio_urubra, family = quasibinomial())
##
## Survey design:
## svydesign(id = ~numer_esc, strata = ~pais, weights = ~round(weight.rec,
## 1), data = brasuru, nest = TRUE)
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -0.414 0.164 -2.53 0.0133 *
## socioecon4cat2-medio-alto 0.636 0.229 2.78 0.0069 **
## socioecon4cat3-medio-bajo 0.969 0.190 5.11 2.2e-06 ***
## socioecon4cat4-bajo 1.466 0.233 6.29 1.7e-08 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for quasibinomial family taken to be 1)
##
## Number of Fisher Scoring iterations: 4
confint(modelo2.bin)
## 2.5 % 97.5 %
## (Intercept) -0.7349 -0.09387
## socioecon4cat2-medio-alto 0.1870 1.08521
## socioecon4cat3-medio-bajo 0.5977 1.34079
## socioecon4cat4-bajo 1.0091 1.92283
exp(modelo2.bin$coefficients)
## (Intercept) socioecon4cat2-medio-alto
## 0.6607 1.8891
## socioecon4cat3-medio-bajo socioecon4cat4-bajo
## 2.6360 4.3318
modelo3.bin <- svyglm(prevoms ~ socioecon3cat, disenio_urubra, family = quasibinomial())
summary(modelo3.bin)
##
## Call:
## svyglm(formula = prevoms ~ socioecon3cat, disenio_urubra, family = quasibinomial())
##
## Survey design:
## svydesign(id = ~numer_esc, strata = ~pais, weights = ~round(weight.rec,
## 1), data = brasuru, nest = TRUE)
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -0.246 0.132 -1.86 0.067 .
## socioecon3cat2-medio 0.683 0.151 4.51 2.2e-05 ***
## socioecon3cat3-bajo 1.298 0.198 6.56 5.2e-09 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for quasibinomial family taken to be 1)
##
## Number of Fisher Scoring iterations: 4
confint(modelo3.bin)
## 2.5 % 97.5 %
## (Intercept) -0.5059 0.01318
## socioecon3cat2-medio 0.3862 0.97989
## socioecon3cat3-bajo 0.9100 1.68580
exp(modelo3.bin$coefficients)
## (Intercept) socioecon3cat2-medio socioecon3cat3-bajo
## 0.7817 1.9799 3.6617
modelo4.bin <- svyglm(prevoms ~ escolmaerecat23cat, disenio_urubra, family = quasibinomial())
summary(modelo4.bin)
##
## Call:
## svyglm(formula = prevoms ~ escolmaerecat23cat, disenio_urubra,
## family = quasibinomial())
##
## Survey design:
## svydesign(id = ~numer_esc, strata = ~pais, weights = ~round(weight.rec,
## 1), data = brasuru, nest = TRUE)
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -0.214 0.131 -1.63 0.10722
## escolmaerecat23cat2-high school 0.550 0.155 3.54 0.00068
## escolmaerecat23cat3-elementary school 0.999 0.172 5.81 1.3e-07
##
## (Intercept)
## escolmaerecat23cat2-high school ***
## escolmaerecat23cat3-elementary school ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for quasibinomial family taken to be 1.003)
##
## Number of Fisher Scoring iterations: 4
confint(modelo4.bin)
## 2.5 % 97.5 %
## (Intercept) -0.4704 0.04332
## escolmaerecat23cat2-high school 0.2452 0.85457
## escolmaerecat23cat3-elementary school 0.6620 1.33616
exp(modelo4.bin$coefficients)
## (Intercept)
## 0.8077
## escolmaerecat23cat2-high school
## 1.7330
## escolmaerecat23cat3-elementary school
## 2.7158
modelo5.bin <- svyglm(prevoms ~ escolmae13cat, disenio_urubra, family = quasibinomial())
summary(modelo5.bin)
##
## Call:
## svyglm(formula = prevoms ~ escolmae13cat, disenio_urubra, family = quasibinomial())
##
## Survey design:
## svydesign(id = ~numer_esc, strata = ~pais, weights = ~round(weight.rec,
## 1), data = brasuru, nest = TRUE)
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -0.471 0.139 -3.40 0.0011 **
## escolmae13cat2-high school 0.530 0.235 2.26 0.0267 *
## escolmae13cat3-Elementary School 1.190 0.176 6.77 2.1e-09 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for quasibinomial family taken to be 1.003)
##
## Number of Fisher Scoring iterations: 4
confint(modelo5.bin)
## 2.5 % 97.5 %
## (Intercept) -0.74290 -0.1998
## escolmae13cat2-high school 0.06988 0.9906
## escolmae13cat3-Elementary School 0.84540 1.5349
exp(modelo5.bin$coefficients)
## (Intercept) escolmae13cat2-high school
## 0.6242 1.6994
## escolmae13cat3-Elementary School
## 3.2876
modelo6.bin <- svyglm(prevoms ~ freqescovacat, disenio_urubra, family = quasibinomial())
summary(modelo6.bin)
##
## Call:
## svyglm(formula = prevoms ~ freqescovacat, disenio_urubra, family = quasibinomial())
##
## Survey design:
## svydesign(id = ~numer_esc, strata = ~pais, weights = ~round(weight.rec,
## 1), data = brasuru, nest = TRUE)
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 0.758 0.123 6.17 2.8e-08 ***
## freqescovacat2-veces al d\xeda -0.402 0.126 -3.19 0.00203 **
## freqescovacat3-veces al d\xeda -0.575 0.146 -3.95 0.00017 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for quasibinomial family taken to be 1.001)
##
## Number of Fisher Scoring iterations: 4
confint(modelo6.bin)
## 2.5 % 97.5 %
## (Intercept) 0.5167 0.9984
## freqescovacat2-veces al d\xeda -0.6489 -0.1551
## freqescovacat3-veces al d\xeda -0.8599 -0.2892
exp(modelo6.bin$coefficients)
## (Intercept) freqescovacat2-veces al d\xeda
## 2.133 0.669
## freqescovacat3-veces al d\xeda
## 0.563
modelo7.bin <- svyglm(prevoms ~ usocreme, disenio_urubra, family = quasibinomial())
summary(modelo7.bin)
##
## Call:
## svyglm(formula = prevoms ~ usocreme, disenio_urubra, family = quasibinomial())
##
## Survey design:
## svydesign(id = ~numer_esc, strata = ~pais, weights = ~round(weight.rec,
## 1), data = brasuru, nest = TRUE)
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 0.347 0.076 4.57 1.8e-05 ***
## usocreme2-No 0.582 0.381 1.53 0.13
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for quasibinomial family taken to be 1.002)
##
## Number of Fisher Scoring iterations: 4
confint(modelo7.bin)
## 2.5 % 97.5 %
## (Intercept) 0.1980 0.4958
## usocreme2-No -0.1653 1.3289
exp(modelo7.bin$coefficients)
## (Intercept) usocreme2-No
## 1.415 1.789
modelo8.bin <- svyglm(prevoms ~ visidentcatonde, disenio_urubra, family = quasibinomial())
summary(modelo8.bin)
##
## Call:
## svyglm(formula = prevoms ~ visidentcatonde, disenio_urubra, family = quasibinomial())
##
## Survey design:
## svydesign(id = ~numer_esc, strata = ~pais, weights = ~round(weight.rec,
## 1), data = brasuru, nest = TRUE)
##
## Coefficients:
## Estimate Std. Error t value
## (Intercept) 0.0447 0.0764 0.59
## visidentcatonde2-Publico 0.8203 0.1138 7.21
## visidentcatonde3-Nunca fue al dentista 0.5164 0.1323 3.90
## Pr(>|t|)
## (Intercept) 0.5601
## visidentcatonde2-Publico 3e-10 ***
## visidentcatonde3-Nunca fue al dentista 0.0002 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for quasibinomial family taken to be 0.9956)
##
## Number of Fisher Scoring iterations: 4
confint(modelo8.bin)
## 2.5 % 97.5 %
## (Intercept) -0.1050 0.1943
## visidentcatonde2-Publico 0.5973 1.0433
## visidentcatonde3-Nunca fue al dentista 0.2570 0.7757
exp(modelo8.bin$coefficients)
## (Intercept)
## 1.046
## visidentcatonde2-Publico
## 2.271
## visidentcatonde3-Nunca fue al dentista
## 1.676
modelo9.bin <- svyglm(prevoms ~ visiquando, disenio_urubra, family = quasibinomial())
summary(modelo9.bin)
##
## Call:
## svyglm(formula = prevoms ~ visiquando, disenio_urubra, family = quasibinomial())
##
## Survey design:
## svydesign(id = ~numer_esc, strata = ~pais, weights = ~round(weight.rec,
## 1), data = brasuru, nest = TRUE)
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 0.30832 0.07775 3.97 0.00016 ***
## visiquando2- m\xe1s de 2 a\xf1os 0.00728 0.15435 0.05 0.96250
## visiquando3- nunca fue 0.26977 0.12670 2.13 0.03636 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for quasibinomial family taken to be 0.9982)
##
## Number of Fisher Scoring iterations: 4
confint(modelo9.bin)
## 2.5 % 97.5 %
## (Intercept) 0.15593 0.4607
## visiquando2- m\xe1s de 2 a\xf1os -0.29525 0.3098
## visiquando3- nunca fue 0.02143 0.5181
exp(modelo9.bin$coefficients)
## (Intercept) visiquando2- m\xe1s de 2 a\xf1os
## 1.361 1.007
## visiquando3- nunca fue
## 1.310
modelo10.bin <- svyglm(prevoms ~ fluorprof, disenio_urubra, family = quasibinomial())
summary(modelo10.bin)
##
## Call:
## svyglm(formula = prevoms ~ fluorprof, disenio_urubra, family = quasibinomial())
##
## Survey design:
## svydesign(id = ~numer_esc, strata = ~pais, weights = ~round(weight.rec,
## 1), data = brasuru, nest = TRUE)
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 0.3020 0.0889 3.40 0.0011 **
## fluorprof2-No 0.1015 0.1242 0.82 0.4165
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for quasibinomial family taken to be 0.9991)
##
## Number of Fisher Scoring iterations: 4
confint(modelo10.bin)
## 2.5 % 97.5 %
## (Intercept) 0.1277 0.4763
## fluorprof2-No -0.1420 0.3449
exp(modelo10.bin$coefficients)
## (Intercept) fluorprof2-No
## 1.353 1.107
modelo11.bin <- svyglm(prevoms ~ isg20 + pais, disenio_urubra, family = quasibinomial())
summary(modelo11.bin)
##
## Call:
## svyglm(formula = prevoms ~ isg20 + pais, disenio_urubra, family = quasibinomial())
##
## Survey design:
## svydesign(id = ~numer_esc, strata = ~pais, weights = ~round(weight.rec,
## 1), data = brasuru, nest = TRUE)
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -0.0299 0.1780 -0.17 0.8670
## isg202- 20 a 60 0.0505 0.1689 0.30 0.7659
## isg203- >=60 0.7062 0.1643 4.30 4.9e-05 ***
## paisUru 0.4274 0.1382 3.09 0.0028 **
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for quasibinomial family taken to be 1.001)
##
## Number of Fisher Scoring iterations: 4
confint(modelo11.bin)
## 2.5 % 97.5 %
## (Intercept) -0.3789 0.3190
## isg202- 20 a 60 -0.2806 0.3816
## isg203- >=60 0.3843 1.0282
## paisUru 0.1564 0.6983
exp(modelo11.bin$coefficients)
## (Intercept) isg202- 20 a 60 isg203- >=60 paisUru
## 0.9705 1.0518 2.0263 1.5332
modelo12.bin <- svyglm(prevoms ~ isg45, disenio_urubra, family = quasibinomial())
summary(modelo12.bin)
##
## Call:
## svyglm(formula = prevoms ~ isg45, disenio_urubra, family = quasibinomial())
##
## Survey design:
## svydesign(id = ~numer_esc, strata = ~pais, weights = ~round(weight.rec,
## 1), data = brasuru, nest = TRUE)
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 0.2430 0.0959 2.53 0.013 *
## isg452- 45 a 60 0.1196 0.1293 0.93 0.358
## isg453- >=60 0.5275 0.1246 4.23 6.2e-05 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for quasibinomial family taken to be 1)
##
## Number of Fisher Scoring iterations: 4
confint(modelo12.bin)
## 2.5 % 97.5 %
## (Intercept) 0.05503 0.4309
## isg452- 45 a 60 -0.13379 0.3730
## isg453- >=60 0.28321 0.7717
exp(modelo12.bin$coefficients)
## (Intercept) isg452- 45 a 60 isg453- >=60
## 1.275 1.127 1.695
modelo13.bin <- svyglm(prevoms ~ tipoesc, disenio_urubra, family = quasibinomial())
summary(modelo13.bin)
##
## Call:
## svyglm(formula = prevoms ~ tipoesc, disenio_urubra, family = quasibinomial())
##
## Survey design:
## svydesign(id = ~numer_esc, strata = ~pais, weights = ~round(weight.rec,
## 1), data = brasuru, nest = TRUE)
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -0.181 0.125 -1.45 0.15
## tipoesc2-Publica 0.725 0.137 5.30 9.9e-07 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for quasibinomial family taken to be 1)
##
## Number of Fisher Scoring iterations: 4
confint(modelo13.bin)
## 2.5 % 97.5 %
## (Intercept) -0.4265 0.06432
## tipoesc2-Publica 0.4568 0.99231
exp(modelo13.bin$coefficients)
## (Intercept) tipoesc2-Publica
## 0.8344 2.0638
#
# load('~/Dropbox/odontologia/maestria_anunziatta/BancoUnidoBrasUru/BRASURU(11102013).RData')
#
# modelo1multi.bin<-svyglm(prevoms~pais+tipoesc+socioecon4cat+escolmae13cat+freqescovacat+visidentcatonde+visiquando+isg45,disenio_urubra,family=quasibinomial())
# summary(modelo1multi.bin) confint(modelo1multi.bin)
# exp(modelo1multi.bin$coefficients)