Simple spatial models for RV, UV, V (Russia)
Libraries
library(tidyverse)
library(dplyr)
library(stargazer)
# geo
library(sp)
library(spdep)
library(rgdal)
library(maptools)
library(rgeos)
library(sf)
library(spatialreg)
# mass retype()
library(hablar)
# multicollinearity test omcdiag imcdiag
library(mctest)
library(GGally)
#tables
library(xtable)
library(texreg)#Specifications for 2017
grp_2017_RV_UV <- log(GRP_2017) ~ Rv_2017 + Uv_2017 + log(Inv_2017) + educ_2017 + Wage_2017
labpr_2017_RV_UV <- LabPr_2017 ~ Rv_2017 + Uv_2017 + log(Inv_2017) + educ_2017 + Wage_2017
emp_2017_RV_UV <- Emp_2017 ~ Rv_2017 + Uv_2017 + log(Inv_2017) + educ_2017 + Wage_2017
unemp_2017_RV_UV <- Unemp_2017 ~ Rv_2017 + Uv_2017 + log(Inv_2017) + educ_2017 + Wage_2017grp_2017_V <- log(GRP_2017) ~ V_2017 + log(Inv_2017) + educ_2017 + Wage_2017
labpr_2017_V <- LabPr_2017 ~ V_2017 + log(Inv_2017) + educ_2017 + Wage_2017
emp_2017_V <- Emp_2017 ~ V_2017 + log(Inv_2017) + educ_2017 + Wage_2017
unemp_2017_V <- Unemp_2017 ~ V_2017 + log(Inv_2017) + educ_2017 + Wage_2017#Spatial Analysis
#downloading spatial data
Russia_shape = readOGR(
'C:/Users/Kolya/Documents/New_studies/GeoAnalisys/DataGeo/Term_paper/Spatial_dataset.shp')## OGR data source with driver: ESRI Shapefile
## Source: "C:\Users\Kolya\Documents\New_studies\GeoAnalisys\DataGeo\Term_paper\Spatial_dataset.shp", layer: "Spatial_dataset"
## with 79 features
## It has 11 fields
## Integer64 fields read as strings: educ_2017 Inv_2017#should be done in shp!
# #our dataframe
# Bad_regions_sp = c("Crimea", "Khanty-Mansiy", "Yamal-Nenets", "Sevastopol'", "Chechnya", "Nenets")
# Russia_shape@data <- subset(Russia_shape@data,!(Russia_shape@data$NAME_1 %in% Bad_regions_sp))
# #drop duplicates
# Russia_shape@data <- Russia_shape@data[!duplicated(Russia_shape@data$NAME_1),]
# #drop useless columns
# Russia_shape@data <- subset(Russia_shape@data, select = -c(1:4, 6:15))
#
# #rename first column
# names(Russia_shape@data)[1] <- 'region'
# names(Russia_shape@data)[9] <- 'V_2017'
# names(Russia_shape@data)[7] <- 'Emp_2017'
# Russia_shape@data
#changing the data type
Russia_shape$educ_2017 <- as.numeric(Russia_shape$educ_2017)
glimpse(Russia_shape$educ_2017)#success## num [1:79] 275 218 199 164 289 255 316 223 253 278 ...Russia_shape$Inv_2017 <- as.numeric(Russia_shape$Inv_2017)
glimpse(Russia_shape$Inv_2017)#success## num [1:79] 45977 37256 240560 185708 144040 ...#Descriptive statistics
Russia_shape@data %>% glimpse()## Rows: 79
## Columns: 11
## $ NAME_1 <chr> "Adygey", "Altay", "Amur", "Arkhangel'sk", "Astrakhan'", "B~
## $ educ_2017 <dbl> 275, 218, 199, 164, 289, 255, 316, 223, 253, 278, 288, 549,~
## $ GRP_2017 <dbl> 109714.5, 545303.0, 299181.0, 759206.5, 442608.8, 1487892.2~
## $ LabPr_2017 <dbl> 103.5, 100.7, 97.8, 103.9, 109.9, 102.9, 103.1, 105.6, 101.~
## $ Unemp_2017 <dbl> 8.8, 6.9, 5.9, 6.4, 7.4, 5.6, 3.9, 4.4, 9.6, 6.6, 5.1, 1.6,~
## $ Wage_2017 <dbl> 101.6, 100.0, 101.1, 98.9, 97.1, 99.2, 99.1, 99.4, 97.9, 97~
## $ Emp_2017 <dbl> 0.0112900000, 0.0153074205, 0.0715903614, 0.0213677966, 0.0~
## $ Inv_2017 <dbl> 45977, 37256, 240560, 185708, 144040, 68532, 91978, 45339, ~
## $ V_2017 <dbl> 2.877239, 3.477131, 3.052880, 3.580351, 3.126376, 3.527909,~
## $ Rv_2017 <dbl> 0.31220840, 0.67319887, 0.55326057, 0.42545317, 0.94392724,~
## $ Uv_2017 <dbl> 2.565030, 2.803932, 2.499619, 3.154898, 2.182449, 2.788336,~#Russia_shape@data$region finally, 79proj4string(Russia_shape) <- CRS("+init=epsg:4326")## Warning in proj4string(obj): CRS object has comment, which is lost in output; in tests, see
## https://cran.r-project.org/web/packages/sp/vignettes/CRS_warnings.htmlRussia.centroids<-gCentroid(Russia_shape,byid=TRUE)
weights <- knn2nb(knearneigh(Russia.centroids, k=8,longlat = TRUE),row.names=Russia_shape$region)
Russia.weights <- nb2listw(weights)#H0 - variable is randomly distributed (no-autocrrelation)
columns = names(Russia_shape@data)[2:length(names(Russia_shape@data))]
Russia_sp_dat <- Russia_shape@data
for(i in columns){
res <- moran.test(Russia_sp_dat[[i]], Russia.weights)
# print(res)
if(res$p.value < 0.01){
print(paste(i,': Spatial correlation detected'))
}
}## [1] "Unemp_2017 : Spatial correlation detected"
## [1] "Emp_2017 : Spatial correlation detected"
## [1] "Inv_2017 : Spatial correlation detected"
## [1] "V_2017 : Spatial correlation detected"
## [1] "Rv_2017 : Spatial correlation detected"#only 5/10 variables appear to be spatially correlated
#Unemp_2017, EmpPerWF_2, Inv_2017, V_2017, Rv_2017for (g in columns){
moran.plot(Russia_shape@data[[g]], Russia.weights, xlab=g)
}
#graphically, all results of tests are correct#Naive spatial models, RV, UV ##GRP, Rv, Uv
attach(Russia_shape@data) #strange ritual
reg_grp_2017_RV_UV <- lm(grp_2017_RV_UV, data = Russia_shape@data)
summary(reg_grp_2017_RV_UV) #%>% xtable()##
## Call:
## lm(formula = grp_2017_RV_UV, data = Russia_shape@data)
##
## Residuals:
## Min 1Q Median 3Q Max
## -1.72792 -0.41127 -0.02262 0.47348 2.00452
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 3.379190 5.153948 0.656 0.5141
## Rv_2017 0.182247 0.151832 1.200 0.2339
## Uv_2017 0.782195 0.172554 4.533 2.23e-05 ***
## log(Inv_2017) 0.693175 0.142246 4.873 6.21e-06 ***
## educ_2017 0.002094 0.001052 1.990 0.0503 .
## Wage_2017 -0.007027 0.051079 -0.138 0.8910
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.7715 on 73 degrees of freedom
## Multiple R-squared: 0.5458, Adjusted R-squared: 0.5147
## F-statistic: 17.55 on 5 and 73 DF, p-value: 2.238e-11#H0 - residuals are randomly distributed (no-autocorrelation)
lm.morantest(reg_grp_2017_RV_UV,Russia.weights)##
## Global Moran I for regression residuals
##
## data:
## model: lm(formula = grp_2017_RV_UV, data = Russia_shape@data)
## weights: Russia.weights
##
## Moran I statistic standard deviate = 6.2213, p-value = 2.466e-10
## alternative hypothesis: greater
## sample estimates:
## Observed Moran I Expectation Variance
## 0.273592461 -0.018646456 0.002206571moran.test(reg_grp_2017_RV_UV$residuals, Russia.weights)##
## Moran I test under randomisation
##
## data: reg_grp_2017_RV_UV$residuals
## weights: Russia.weights
##
## Moran I statistic standard deviate = 5.7772, p-value = 3.797e-09
## alternative hypothesis: greater
## sample estimates:
## Moran I statistic Expectation Variance
## 0.27359246 -0.01282051 0.00245779moran.plot(reg_grp_2017_RV_UV$residuals, Russia.weights)
# residuals are correlated according to the both test and graphlm.LMtests(reg_grp_2017_RV_UV, listw = Russia.weights, test="all", zero.policy=TRUE)##
## Lagrange multiplier diagnostics for spatial dependence
##
## data:
## model: lm(formula = grp_2017_RV_UV, data = Russia_shape@data)
## weights: Russia.weights
##
## LMerr = 26.695, df = 1, p-value = 2.383e-07
##
##
## Lagrange multiplier diagnostics for spatial dependence
##
## data:
## model: lm(formula = grp_2017_RV_UV, data = Russia_shape@data)
## weights: Russia.weights
##
## LMlag = 7.5941, df = 1, p-value = 0.005856
##
##
## Lagrange multiplier diagnostics for spatial dependence
##
## data:
## model: lm(formula = grp_2017_RV_UV, data = Russia_shape@data)
## weights: Russia.weights
##
## RLMerr = 21.101, df = 1, p-value = 4.358e-06
##
##
## Lagrange multiplier diagnostics for spatial dependence
##
## data:
## model: lm(formula = grp_2017_RV_UV, data = Russia_shape@data)
## weights: Russia.weights
##
## RLMlag = 2.0002, df = 1, p-value = 0.1573
##
##
## Lagrange multiplier diagnostics for spatial dependence
##
## data:
## model: lm(formula = grp_2017_RV_UV, data = Russia_shape@data)
## weights: Russia.weights
##
## SARMA = 28.695, df = 2, p-value = 5.875e-07# Analitics
# LMerr = 26.695, df = 1, p-value = 2.383e-07 => significant
# LMlag = 7.5941, df = 1, p-value = 0.005856 => significant
# conclusion => moving to Robust LM Diagnostics
# RLMerr = 21.101, df = 1, p-value = 4.358e-06
# RLMlag = 2.0002, df = 1, p-value = 0.1573
# p-value: RLMlag > RLMerr => RLMerr is robust
# #SEM - spatial error model
# y = X beta + rho W1 y + u
# u = lambda W2 u + e#building an actual model
SEM_GRP_RV_UV <- errorsarlm(grp_2017_RV_UV, listw = Russia.weights, zero.policy=TRUE)
sum_GRP_RV_UV <- summary(SEM_GRP_RV_UV)
sum_GRP_RV_UV$Coef[, c(1,4)] %>% xtable()# print("NOX")
# print(sum_b$Coef[c(11),])
#comparing our spatial model with the original one
logLik(SEM_GRP_RV_UV)## 'log Lik.' -77.28472 (df=8)logLik(reg_grp_2017_RV_UV)## 'log Lik.' -88.48648 (df=7)AIC(SEM_GRP_RV_UV)## [1] 170.5694AIC(reg_grp_2017_RV_UV)## [1] 190.973#AIC - the bigger, the better the model
#log-likelihood - the smaller, the better the model
#SER is better than OLS in this case##LabPr, Rv, Uv
#strange ritual
reg_labpr_2017_RV_UV <- lm(labpr_2017_RV_UV, data = Russia_shape@data)
summary(reg_labpr_2017_RV_UV) #%>% xtable()##
## Call:
## lm(formula = labpr_2017_RV_UV, data = Russia_shape@data)
##
## Residuals:
## Min 1Q Median 3Q Max
## -7.4942 -1.6520 -0.0417 1.5016 10.0609
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 134.432680 19.052399 7.056 8.2e-10 ***
## Rv_2017 -0.118557 0.561273 -0.211 0.833
## Uv_2017 0.517382 0.637873 0.811 0.420
## log(Inv_2017) -0.244087 0.525836 -0.464 0.644
## educ_2017 -0.003996 0.003890 -1.027 0.308
## Wage_2017 -0.298336 0.188822 -1.580 0.118
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 2.852 on 73 degrees of freedom
## Multiple R-squared: 0.0531, Adjusted R-squared: -0.01176
## F-statistic: 0.8187 on 5 and 73 DF, p-value: 0.5403#H0 - residuals are randomly distributed (no-autocorrelation)
lm.morantest(reg_labpr_2017_RV_UV,Russia.weights)##
## Global Moran I for regression residuals
##
## data:
## model: lm(formula = labpr_2017_RV_UV, data = Russia_shape@data)
## weights: Russia.weights
##
## Moran I statistic standard deviate = -0.46781, p-value = 0.68
## alternative hypothesis: greater
## sample estimates:
## Observed Moran I Expectation Variance
## -0.040621347 -0.018646456 0.002206571moran.test(reg_labpr_2017_RV_UV$residuals, Russia.weights)##
## Moran I test under randomisation
##
## data: reg_labpr_2017_RV_UV$residuals
## weights: Russia.weights
##
## Moran I statistic standard deviate = -0.56841, p-value = 0.7151
## alternative hypothesis: greater
## sample estimates:
## Moran I statistic Expectation Variance
## -0.040621347 -0.012820513 0.002392169moran.plot(reg_labpr_2017_RV_UV$residuals, Russia.weights)
# residuals are not correlated according to the both test and graphlm.LMtests(reg_labpr_2017_RV_UV, listw = Russia.weights, test="all", zero.policy=TRUE)##
## Lagrange multiplier diagnostics for spatial dependence
##
## data:
## model: lm(formula = labpr_2017_RV_UV, data = Russia_shape@data)
## weights: Russia.weights
##
## LMerr = 0.58847, df = 1, p-value = 0.443
##
##
## Lagrange multiplier diagnostics for spatial dependence
##
## data:
## model: lm(formula = labpr_2017_RV_UV, data = Russia_shape@data)
## weights: Russia.weights
##
## LMlag = 0.55568, df = 1, p-value = 0.456
##
##
## Lagrange multiplier diagnostics for spatial dependence
##
## data:
## model: lm(formula = labpr_2017_RV_UV, data = Russia_shape@data)
## weights: Russia.weights
##
## RLMerr = 0.040606, df = 1, p-value = 0.8403
##
##
## Lagrange multiplier diagnostics for spatial dependence
##
## data:
## model: lm(formula = labpr_2017_RV_UV, data = Russia_shape@data)
## weights: Russia.weights
##
## RLMlag = 0.0078192, df = 1, p-value = 0.9295
##
##
## Lagrange multiplier diagnostics for spatial dependence
##
## data:
## model: lm(formula = labpr_2017_RV_UV, data = Russia_shape@data)
## weights: Russia.weights
##
## SARMA = 0.59629, df = 2, p-value = 0.7422# Analitics
# LMerr = 26.695, df = 1, p-value = 2.383e-07 => non -significant
# LMlag = 7.5941, df = 1, p-value = 0.005856 => non -significant
# conclusion => sticking to OLS##Emp, Rv, Uv
#strange ritual
reg_emp_2017_RV_UV <- lm(emp_2017_RV_UV, data = Russia_shape@data)
summary(reg_emp_2017_RV_UV) #%>% xtable()##
## Call:
## lm(formula = emp_2017_RV_UV, data = Russia_shape@data)
##
## Residuals:
## Min 1Q Median 3Q Max
## -0.018458 -0.005822 -0.001369 0.002788 0.052274
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -7.338e-02 7.682e-02 -0.955 0.342609
## Rv_2017 -7.878e-05 2.263e-03 -0.035 0.972324
## Uv_2017 6.357e-03 2.572e-03 2.472 0.015775 *
## log(Inv_2017) 7.294e-03 2.120e-03 3.440 0.000964 ***
## educ_2017 -3.141e-05 1.568e-05 -2.003 0.048922 *
## Wage_2017 2.450e-05 7.614e-04 0.032 0.974413
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.0115 on 73 degrees of freedom
## Multiple R-squared: 0.243, Adjusted R-squared: 0.1911
## F-statistic: 4.686 on 5 and 73 DF, p-value: 0.000913#H0 - residuals are randomly distributed (no-autocorrelation)
lm.morantest(reg_emp_2017_RV_UV,Russia.weights)##
## Global Moran I for regression residuals
##
## data:
## model: lm(formula = emp_2017_RV_UV, data = Russia_shape@data)
## weights: Russia.weights
##
## Moran I statistic standard deviate = 5.7608, p-value = 4.185e-09
## alternative hypothesis: greater
## sample estimates:
## Observed Moran I Expectation Variance
## 0.251964083 -0.018646456 0.002206571moran.test(reg_emp_2017_RV_UV$residuals, Russia.weights)##
## Moran I test under randomisation
##
## data: reg_emp_2017_RV_UV$residuals
## weights: Russia.weights
##
## Moran I statistic standard deviate = 5.6529, p-value = 7.888e-09
## alternative hypothesis: greater
## sample estimates:
## Moran I statistic Expectation Variance
## 0.251964083 -0.012820513 0.002194021moran.plot(reg_emp_2017_RV_UV$residuals, Russia.weights)
# residuals are correlated according to the both test and graphlm.LMtests(reg_emp_2017_RV_UV, listw = Russia.weights, test="all", zero.policy=TRUE)##
## Lagrange multiplier diagnostics for spatial dependence
##
## data:
## model: lm(formula = emp_2017_RV_UV, data = Russia_shape@data)
## weights: Russia.weights
##
## LMerr = 22.641, df = 1, p-value = 1.953e-06
##
##
## Lagrange multiplier diagnostics for spatial dependence
##
## data:
## model: lm(formula = emp_2017_RV_UV, data = Russia_shape@data)
## weights: Russia.weights
##
## LMlag = 40.163, df = 1, p-value = 2.336e-10
##
##
## Lagrange multiplier diagnostics for spatial dependence
##
## data:
## model: lm(formula = emp_2017_RV_UV, data = Russia_shape@data)
## weights: Russia.weights
##
## RLMerr = 1.7098, df = 1, p-value = 0.191
##
##
## Lagrange multiplier diagnostics for spatial dependence
##
## data:
## model: lm(formula = emp_2017_RV_UV, data = Russia_shape@data)
## weights: Russia.weights
##
## RLMlag = 19.232, df = 1, p-value = 1.158e-05
##
##
## Lagrange multiplier diagnostics for spatial dependence
##
## data:
## model: lm(formula = emp_2017_RV_UV, data = Russia_shape@data)
## weights: Russia.weights
##
## SARMA = 41.873, df = 2, p-value = 8.081e-10# Analitics
# LMerr = 22.641, df = 1, p-value = 1.953e-06 => significant
# LMlag = 40.163, df = 1, p-value = 2.336e-10 => significant
# conclusion => moving to Robust LM Diagnostics
# RLMerr = 1.7098, df = 1, p-value = 0.191
# RLMlag = 19.232, df = 1, p-value = 1.158e-05
# p-value: RLMlag < RLMerr => RLMlag is robust
# SAR - spatial autoregressive model
# y = X beta + rho W1 y + u#building an actual model
SAR_Emp_RV_UV <- lagsarlm(emp_2017_RV_UV, listw = Russia.weights, zero.policy=TRUE)
sum_Emp_RV_UV <- summary(SAR_Emp_RV_UV)
sum_Emp_RV_UV$Coef[, c(1,4)] %>% xtable()# print("NOX")
# print(sum_b$Coef[c(11),])
#comparing our spatial model with the original one
logLik(SAR_Emp_RV_UV)## 'log Lik.' 254.7317 (df=8)logLik(reg_emp_2017_RV_UV)## 'log Lik.' 243.7909 (df=7)AIC(SAR_Emp_RV_UV)## [1] -493.4633AIC(reg_emp_2017_RV_UV)## [1] -473.5818#AIC - the bigger, the better the model
#log-likelihood - the smaller, the better the model
#SAR is better than OLS in this case##Unemp, Rv, Uv
#strange ritual
reg_unemp_2017_RV_UV <- lm(unemp_2017_RV_UV, data = Russia_shape@data)
summary(reg_unemp_2017_RV_UV) #%>% xtable()##
## Call:
## lm(formula = unemp_2017_RV_UV, data = Russia_shape@data)
##
## Residuals:
## Min 1Q Median 3Q Max
## -3.9400 -2.0770 -0.5294 1.4176 17.1003
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 40.772412 21.506862 1.896 0.06195 .
## Rv_2017 0.515684 0.633580 0.814 0.41834
## Uv_2017 -1.102269 0.720048 -1.531 0.13013
## log(Inv_2017) -1.928983 0.593578 -3.250 0.00175 **
## educ_2017 -0.006644 0.004391 -1.513 0.13453
## Wage_2017 -0.085552 0.213148 -0.401 0.68932
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 3.22 on 73 degrees of freedom
## Multiple R-squared: 0.2246, Adjusted R-squared: 0.1715
## F-statistic: 4.23 on 5 and 73 DF, p-value: 0.001969#H0 - residuals are randomly distributed (no-autocorrelation)
lm.morantest(reg_unemp_2017_RV_UV,Russia.weights)##
## Global Moran I for regression residuals
##
## data:
## model: lm(formula = unemp_2017_RV_UV, data = Russia_shape@data)
## weights: Russia.weights
##
## Moran I statistic standard deviate = 7.5185, p-value = 2.77e-14
## alternative hypothesis: greater
## sample estimates:
## Observed Moran I Expectation Variance
## 0.334530629 -0.018646456 0.002206571moran.test(reg_unemp_2017_RV_UV$residuals, Russia.weights)##
## Moran I test under randomisation
##
## data: reg_unemp_2017_RV_UV$residuals
## weights: Russia.weights
##
## Moran I statistic standard deviate = 7.5524, p-value = 2.137e-14
## alternative hypothesis: greater
## sample estimates:
## Moran I statistic Expectation Variance
## 0.334530629 -0.012820513 0.002115296moran.plot(reg_unemp_2017_RV_UV$residuals, Russia.weights)
# residuals are correlated according to the both test and graphlm.LMtests(reg_unemp_2017_RV_UV, listw = Russia.weights, test="all", zero.policy=TRUE)##
## Lagrange multiplier diagnostics for spatial dependence
##
## data:
## model: lm(formula = unemp_2017_RV_UV, data = Russia_shape@data)
## weights: Russia.weights
##
## LMerr = 39.911, df = 1, p-value = 2.659e-10
##
##
## Lagrange multiplier diagnostics for spatial dependence
##
## data:
## model: lm(formula = unemp_2017_RV_UV, data = Russia_shape@data)
## weights: Russia.weights
##
## LMlag = 49.776, df = 1, p-value = 1.723e-12
##
##
## Lagrange multiplier diagnostics for spatial dependence
##
## data:
## model: lm(formula = unemp_2017_RV_UV, data = Russia_shape@data)
## weights: Russia.weights
##
## RLMerr = 0.17875, df = 1, p-value = 0.6725
##
##
## Lagrange multiplier diagnostics for spatial dependence
##
## data:
## model: lm(formula = unemp_2017_RV_UV, data = Russia_shape@data)
## weights: Russia.weights
##
## RLMlag = 10.044, df = 1, p-value = 0.001528
##
##
## Lagrange multiplier diagnostics for spatial dependence
##
## data:
## model: lm(formula = unemp_2017_RV_UV, data = Russia_shape@data)
## weights: Russia.weights
##
## SARMA = 49.955, df = 2, p-value = 1.42e-11# Analitics
# LMerr = 39.911, df = 1, p-value = 2.659e-10 => significant
# LMlag = 49.776, df = 1, p-value = 1.723e-12 => significant
# conclusion => moving to Robust LM Diagnostics
# RLMerr = 0.17875, df = 1, p-value = 0.6725
# RLMlag = 10.044, df = 1, p-value = 0.001528
# p-value: RLMlag < RLMerr => RLMlag is robust
# SAR - spatial autoregressive model
# y = X beta + rho W1 y + u#building an actual model
SAR_Unemp_RV_UV <- lagsarlm(unemp_2017_RV_UV, listw = Russia.weights, zero.policy=TRUE)
sum_Unemp_RV_UV <- summary(SAR_Unemp_RV_UV)
sum_Unemp_RV_UV$Coef[, c(1,4)] %>% xtable()# print("NOX")
# print(sum_b$Coef[c(11),])
#comparing our spatial model with the original one
logLik(SAR_Unemp_RV_UV)## 'log Lik.' -186.7775 (df=8)logLik(reg_unemp_2017_RV_UV)## 'log Lik.' -201.3466 (df=7)AIC(SAR_Unemp_RV_UV)## [1] 389.5549AIC(reg_unemp_2017_RV_UV)## [1] 416.6932#AIC - the bigger, the better the model
#log-likelihood - the smaller, the better the model
#SAR is better than OLS in this case#Stars
library(gtsummary)## Warning: пакет 'gtsummary' был собран под R версии 4.1.3packageVersion("gtsummary")## [1] '1.5.2'style_pvalue_stars <- function(x) {
dplyr::case_when(
x < 0.001 ~ paste0(style_pvalue(x), "***"),
x < 0.01 ~ paste0(style_pvalue(x), "**"),
x < 0.05 ~ paste0(style_pvalue(x), "*"),
TRUE ~ style_pvalue(x)
)
}#Naive spatial models, V ##GRP, V
# attach(Russia_shape@data) #strange ritual
reg_grp_2017_V <- lm(grp_2017_V, data = Russia_shape@data)
summary(reg_grp_2017_V) #%>% xtable()##
## Call:
## lm(formula = grp_2017_V, data = Russia_shape@data)
##
## Residuals:
## Min 1Q Median 3Q Max
## -1.78289 -0.51180 0.04225 0.40835 2.28803
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 4.856058 5.263340 0.923 0.3592
## V_2017 0.452125 0.100286 4.508 2.40e-05 ***
## log(Inv_2017) 0.715093 0.146065 4.896 5.58e-06 ***
## educ_2017 0.003009 0.001004 2.997 0.0037 **
## Wage_2017 -0.020112 0.052245 -0.385 0.7014
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.794 on 74 degrees of freedom
## Multiple R-squared: 0.5124, Adjusted R-squared: 0.4861
## F-statistic: 19.44 on 4 and 74 DF, p-value: 5.728e-11#H0 - residuals are randomly distributed (no-autocorrelation)
lm.morantest(reg_grp_2017_V,Russia.weights)##
## Global Moran I for regression residuals
##
## data:
## model: lm(formula = grp_2017_V, data = Russia_shape@data)
## weights: Russia.weights
##
## Moran I statistic standard deviate = 6.4592, p-value = 5.263e-11
## alternative hypothesis: greater
## sample estimates:
## Observed Moran I Expectation Variance
## 0.2860602 -0.0186465 0.0022254moran.test(reg_grp_2017_V$residuals, Russia.weights)##
## Moran I test under randomisation
##
## data: reg_grp_2017_V$residuals
## weights: Russia.weights
##
## Moran I statistic standard deviate = 6.0395, p-value = 7.732e-10
## alternative hypothesis: greater
## sample estimates:
## Moran I statistic Expectation Variance
## 0.286060226 -0.012820513 0.002449061moran.plot(reg_grp_2017_V$residuals, Russia.weights)
# residuals are correlated according to the both test and graphlm.LMtests(reg_grp_2017_V, listw = Russia.weights, test="all", zero.policy=TRUE)##
## Lagrange multiplier diagnostics for spatial dependence
##
## data:
## model: lm(formula = grp_2017_V, data = Russia_shape@data)
## weights: Russia.weights
##
## LMerr = 29.183, df = 1, p-value = 6.585e-08
##
##
## Lagrange multiplier diagnostics for spatial dependence
##
## data:
## model: lm(formula = grp_2017_V, data = Russia_shape@data)
## weights: Russia.weights
##
## LMlag = 8.4352, df = 1, p-value = 0.00368
##
##
## Lagrange multiplier diagnostics for spatial dependence
##
## data:
## model: lm(formula = grp_2017_V, data = Russia_shape@data)
## weights: Russia.weights
##
## RLMerr = 23.731, df = 1, p-value = 1.108e-06
##
##
## Lagrange multiplier diagnostics for spatial dependence
##
## data:
## model: lm(formula = grp_2017_V, data = Russia_shape@data)
## weights: Russia.weights
##
## RLMlag = 2.9827, df = 1, p-value = 0.08416
##
##
## Lagrange multiplier diagnostics for spatial dependence
##
## data:
## model: lm(formula = grp_2017_V, data = Russia_shape@data)
## weights: Russia.weights
##
## SARMA = 32.166, df = 2, p-value = 1.036e-07# Analitics
# LMerr = 29.183, df = 1, p-value = 6.585e-08 => significant
# LMlag = 8.4352, df = 1, p-value = 0.00368 => significant
# conclusion => moving to Robust LM Diagnostics
# RLMerr = 23.731, df = 1, p-value = 1.108e-06
# RLMlag = 2.9827, df = 1, p-value = 0.08416
# p-value: RLMlag > RLMerr => RLMerr is robust
# #SEM - spatial error model
# y = X beta + rho W1 y + u
# u = lambda W2 u + e#building an actual model
SEM_GRP_V <- errorsarlm(grp_2017_V, listw = Russia.weights, zero.policy=TRUE)
sum_GRP_V <- summary(SEM_GRP_V)
sum_GRP_V$Coef[, c(1,4)] %>% xtable()# print("NOX")
# print(sum_b$Coef[c(11),])
#comparing our spatial model with the original one
logLik(SEM_GRP_V)## 'log Lik.' -79.90729 (df=7)logLik(reg_grp_2017_V)## 'log Lik.' -91.29149 (df=6)AIC(SEM_GRP_V)## [1] 173.8146AIC(reg_grp_2017_V)## [1] 194.583#AIC - the bigger, the better the model
#log-likelihood - the smaller, the better the model
#SER is better than OLS in this case##LabPr, V
#strange ritual
reg_labpr_2017_V <- lm(labpr_2017_V, data = Russia_shape@data)
summary(reg_labpr_2017_V) #%>% xtable()##
## Call:
## lm(formula = labpr_2017_V, data = Russia_shape@data)
##
## Residuals:
## Min 1Q Median 3Q Max
## -7.6667 -1.6670 0.0114 1.3682 10.2422
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 135.998145 18.834791 7.221 3.81e-10 ***
## V_2017 0.167511 0.358870 0.467 0.6420
## log(Inv_2017) -0.220854 0.522690 -0.423 0.6739
## educ_2017 -0.003026 0.003592 -0.842 0.4022
## Wage_2017 -0.312205 0.186956 -1.670 0.0992 .
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 2.841 on 74 degrees of freedom
## Multiple R-squared: 0.04737, Adjusted R-squared: -0.004128
## F-statistic: 0.9198 on 4 and 74 DF, p-value: 0.4571#H0 - residuals are randomly distributed (no-autocorrelation)
lm.morantest(reg_labpr_2017_V,Russia.weights)##
## Global Moran I for regression residuals
##
## data:
## model: lm(formula = labpr_2017_V, data = Russia_shape@data)
## weights: Russia.weights
##
## Moran I statistic standard deviate = -0.31217, p-value = 0.6225
## alternative hypothesis: greater
## sample estimates:
## Observed Moran I Expectation Variance
## -0.03337275 -0.01864650 0.00222540moran.test(reg_labpr_2017_V$residuals, Russia.weights)##
## Moran I test under randomisation
##
## data: reg_labpr_2017_V$residuals
## weights: Russia.weights
##
## Moran I statistic standard deviate = -0.42047, p-value = 0.6629
## alternative hypothesis: greater
## sample estimates:
## Moran I statistic Expectation Variance
## -0.033372746 -0.012820513 0.002389148moran.plot(reg_labpr_2017_V$residuals, Russia.weights)
# residuals are not correlated according to the both test and graphlm.LMtests(reg_labpr_2017_V, listw = Russia.weights, test="all", zero.policy=TRUE)##
## Lagrange multiplier diagnostics for spatial dependence
##
## data:
## model: lm(formula = labpr_2017_V, data = Russia_shape@data)
## weights: Russia.weights
##
## LMerr = 0.39719, df = 1, p-value = 0.5285
##
##
## Lagrange multiplier diagnostics for spatial dependence
##
## data:
## model: lm(formula = labpr_2017_V, data = Russia_shape@data)
## weights: Russia.weights
##
## LMlag = 0.40348, df = 1, p-value = 0.5253
##
##
## Lagrange multiplier diagnostics for spatial dependence
##
## data:
## model: lm(formula = labpr_2017_V, data = Russia_shape@data)
## weights: Russia.weights
##
## RLMerr = 2.9529e-05, df = 1, p-value = 0.9957
##
##
## Lagrange multiplier diagnostics for spatial dependence
##
## data:
## model: lm(formula = labpr_2017_V, data = Russia_shape@data)
## weights: Russia.weights
##
## RLMlag = 0.0063212, df = 1, p-value = 0.9366
##
##
## Lagrange multiplier diagnostics for spatial dependence
##
## data:
## model: lm(formula = labpr_2017_V, data = Russia_shape@data)
## weights: Russia.weights
##
## SARMA = 0.40351, df = 2, p-value = 0.8173# Analitics
# LMerr = 26.695, df = 1, p-value = 2.383e-07 => non -significant
# LMlag = 7.5941, df = 1, p-value = 0.005856 => non -significant
# conclusion => sticking to OLS##Emp, V
#strange ritual
reg_emp_2017_V <- lm(emp_2017_V, data = Russia_shape@data)
summary(reg_emp_2017_V) #%>% xtable()##
## Call:
## lm(formula = emp_2017_V, data = Russia_shape@data)
##
## Residuals:
## Min 1Q Median 3Q Max
## -0.016307 -0.005385 -0.001907 0.002592 0.054110
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -5.754e-02 7.714e-02 -0.746 0.45812
## V_2017 2.816e-03 1.470e-03 1.916 0.05922 .
## log(Inv_2017) 7.529e-03 2.141e-03 3.517 0.00075 ***
## educ_2017 -2.160e-05 1.471e-05 -1.468 0.14636
## Wage_2017 -1.159e-04 7.657e-04 -0.151 0.88014
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.01164 on 74 degrees of freedom
## Multiple R-squared: 0.2141, Adjusted R-squared: 0.1716
## F-statistic: 5.04 on 4 and 74 DF, p-value: 0.001199#H0 - residuals are randomly distributed (no-autocorrelation)
lm.morantest(reg_emp_2017_V,Russia.weights)##
## Global Moran I for regression residuals
##
## data:
## model: lm(formula = emp_2017_V, data = Russia_shape@data)
## weights: Russia.weights
##
## Moran I statistic standard deviate = 5.805, p-value = 3.219e-09
## alternative hypothesis: greater
## sample estimates:
## Observed Moran I Expectation Variance
## 0.2551977 -0.0186465 0.0022254moran.test(reg_emp_2017_V$residuals, Russia.weights)##
## Moran I test under randomisation
##
## data: reg_emp_2017_V$residuals
## weights: Russia.weights
##
## Moran I statistic standard deviate = 5.7524, p-value = 4.4e-09
## alternative hypothesis: greater
## sample estimates:
## Moran I statistic Expectation Variance
## 0.255197663 -0.012820513 0.002170878moran.plot(reg_emp_2017_V$residuals, Russia.weights)
# residuals are correlated according to the both test and graphlm.LMtests(reg_emp_2017_V, listw = Russia.weights, test="all", zero.policy=TRUE)##
## Lagrange multiplier diagnostics for spatial dependence
##
## data:
## model: lm(formula = emp_2017_V, data = Russia_shape@data)
## weights: Russia.weights
##
## LMerr = 23.226, df = 1, p-value = 1.441e-06
##
##
## Lagrange multiplier diagnostics for spatial dependence
##
## data:
## model: lm(formula = emp_2017_V, data = Russia_shape@data)
## weights: Russia.weights
##
## LMlag = 41.312, df = 1, p-value = 1.298e-10
##
##
## Lagrange multiplier diagnostics for spatial dependence
##
## data:
## model: lm(formula = emp_2017_V, data = Russia_shape@data)
## weights: Russia.weights
##
## RLMerr = 2.764, df = 1, p-value = 0.09641
##
##
## Lagrange multiplier diagnostics for spatial dependence
##
## data:
## model: lm(formula = emp_2017_V, data = Russia_shape@data)
## weights: Russia.weights
##
## RLMlag = 20.85, df = 1, p-value = 4.966e-06
##
##
## Lagrange multiplier diagnostics for spatial dependence
##
## data:
## model: lm(formula = emp_2017_V, data = Russia_shape@data)
## weights: Russia.weights
##
## SARMA = 44.076, df = 2, p-value = 2.685e-10# Analitics
# LMerr = 23.226, df = 1, p-value = 1.441e-06 => significant
# LMlag = 20.85, df = 1, p-value = 1.298e-10 => significant
# conclusion => moving to Robust LM Diagnostics
# RLMerr = 2.764, df = 1, p-value = 0.09641
# RLMlag = 19.232, df = 1, p-value = 4.966e-06
# p-value: RLMlag < RLMerr => RLMlag is robust
# SAR - spatial autoregressive model
# y = X beta + rho W1 y + u#building an actual model
SAR_Emp_V <- lagsarlm(emp_2017_V, listw = Russia.weights, zero.policy=TRUE)
sum_Emp_V <- summary(SAR_Emp_V)
sum_Emp_V$Coef[, c(1,4)] %>% xtable()# print("NOX")
# print(sum_b$Coef[c(11),])
#comparing our spatial model with the original one
logLik(SAR_Emp_V)## 'log Lik.' 253.3943 (df=7)logLik(reg_emp_2017_V)## 'log Lik.' 242.313 (df=6)AIC(SAR_Emp_V)## [1] -492.7886AIC(reg_emp_2017_V)## [1] -472.626#AIC - the bigger, the better the model
#log-likelihood - the smaller, the better the model
#SAR is better than OLS in this case##Unemp, V
#strange ritual
reg_unemp_2017_V <- lm(unemp_2017_V, data = Russia_shape@data)
summary(reg_unemp_2017_V) #%>% xtable()##
## Call:
## lm(formula = unemp_2017_V, data = Russia_shape@data)
##
## Residuals:
## Min 1Q Median 3Q Max
## -4.0931 -1.9198 -0.4699 1.1838 17.5392
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 36.789563 21.520500 1.710 0.09155 .
## V_2017 -0.212129 0.410043 -0.517 0.60647
## log(Inv_2017) -1.988094 0.597222 -3.329 0.00136 **
## educ_2017 -0.009111 0.004104 -2.220 0.02950 *
## Wage_2017 -0.050265 0.213615 -0.235 0.81462
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 3.247 on 74 degrees of freedom
## Multiple R-squared: 0.2008, Adjusted R-squared: 0.1576
## F-statistic: 4.648 on 4 and 74 DF, p-value: 0.002108#H0 - residuals are randomly distributed (no-autocorrelation)
lm.morantest(reg_unemp_2017_V,Russia.weights)##
## Global Moran I for regression residuals
##
## data:
## model: lm(formula = unemp_2017_V, data = Russia_shape@data)
## weights: Russia.weights
##
## Moran I statistic standard deviate = 8.3878, p-value < 2.2e-16
## alternative hypothesis: greater
## sample estimates:
## Observed Moran I Expectation Variance
## 0.3770416 -0.0186465 0.0022254moran.test(reg_unemp_2017_V$residuals, Russia.weights)##
## Moran I test under randomisation
##
## data: reg_unemp_2017_V$residuals
## weights: Russia.weights
##
## Moran I statistic standard deviate = 8.4991, p-value < 2.2e-16
## alternative hypothesis: greater
## sample estimates:
## Moran I statistic Expectation Variance
## 0.377041625 -0.012820513 0.002104124moran.plot(reg_unemp_2017_V$residuals, Russia.weights)
# residuals are correlated according to the both test and graphlm.LMtests(reg_unemp_2017_V, listw = Russia.weights, test="all", zero.policy=TRUE)##
## Lagrange multiplier diagnostics for spatial dependence
##
## data:
## model: lm(formula = unemp_2017_V, data = Russia_shape@data)
## weights: Russia.weights
##
## LMerr = 50.698, df = 1, p-value = 1.077e-12
##
##
## Lagrange multiplier diagnostics for spatial dependence
##
## data:
## model: lm(formula = unemp_2017_V, data = Russia_shape@data)
## weights: Russia.weights
##
## LMlag = 56.152, df = 1, p-value = 6.706e-14
##
##
## Lagrange multiplier diagnostics for spatial dependence
##
## data:
## model: lm(formula = unemp_2017_V, data = Russia_shape@data)
## weights: Russia.weights
##
## RLMerr = 0.068266, df = 1, p-value = 0.7939
##
##
## Lagrange multiplier diagnostics for spatial dependence
##
## data:
## model: lm(formula = unemp_2017_V, data = Russia_shape@data)
## weights: Russia.weights
##
## RLMlag = 5.5219, df = 1, p-value = 0.01878
##
##
## Lagrange multiplier diagnostics for spatial dependence
##
## data:
## model: lm(formula = unemp_2017_V, data = Russia_shape@data)
## weights: Russia.weights
##
## SARMA = 56.22, df = 2, p-value = 6.193e-13# Analitics
# LMerr = 50.698, df = 1, p-value = 1.077e-12 => significant
# LMlag = 56.152, df = 1, p-value = 6.706e-14 => significant
# conclusion => moving to Robust LM Diagnostics
# RLMerr = 0.068266, df = 1, p-value = 0.7939
# RLMlag = 5.5219, df = 1, p-value =0.01878
# p-value: RLMlag < RLMerr => RLMlag is robust
# SAR - spatial autoregressive model
# y = X beta + rho W1 y + u#building an actual model
SAR_Unemp_V <- lagsarlm(unemp_2017_V, listw = Russia.weights, zero.policy=TRUE)
sum_Unemp_V <- summary(SAR_Unemp_V)
sum_Unemp_V$Coef[, c(1,4)] %>% xtable()# print("NOX")
# print(sum_b$Coef[c(11),])
#comparing our spatial model with the original one
logLik(SAR_Unemp_V)## 'log Lik.' -186.9628 (df=7)logLik(reg_unemp_2017_V)## 'log Lik.' -202.5425 (df=6)AIC(SAR_Unemp_V)## [1] 387.9257AIC(reg_unemp_2017_V)## [1] 417.0849#AIC - the bigger, the better the model
#log-likelihood - the smaller, the better the model
#SAR is better than OLS in this case#Right tables, RV, UV ## SEM, RV, UV
SEM_GRP_RV_UV <- errorsarlm(grp_2017_RV_UV, listw = Russia.weights, zero.policy=TRUE)
sum_GRP_RV_UV <- summary(SEM_GRP_RV_UV)
sum_GRP_RV_UV##
## Call:errorsarlm(formula = grp_2017_RV_UV, listw = Russia.weights,
## zero.policy = TRUE)
##
## Residuals:
## Min 1Q Median 3Q Max
## -1.268640 -0.344029 -0.033312 0.381603 1.579997
##
## Type: error
## Coefficients: (asymptotic standard errors)
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -5.12308318 4.07349607 -1.2577 0.208514
## Rv_2017 0.37582518 0.12838078 2.9274 0.003418
## Uv_2017 0.84269512 0.13450702 6.2651 3.727e-10
## log(Inv_2017) 0.86691086 0.12287194 7.0554 1.721e-12
## educ_2017 0.00076803 0.00083658 0.9180 0.358593
## Wage_2017 0.05866649 0.03977885 1.4748 0.140262
##
## Lambda: 0.70158, LR test value: 22.404, p-value: 2.2097e-06
## Asymptotic standard error: 0.10163
## z-value: 6.9032, p-value: 5.0833e-12
## Wald statistic: 47.655, p-value: 5.0833e-12
##
## Log likelihood: -77.28472 for error model
## ML residual variance (sigma squared): 0.38385, (sigma: 0.61956)
## Number of observations: 79
## Number of parameters estimated: 8
## AIC: 170.57, (AIC for lm: 190.97)# sum_GRP_RV_UV$Coef[, c(1,4)] %>% xtable()SEM_LabPr_RV_UV <- errorsarlm(labpr_2017_RV_UV, listw = Russia.weights, zero.policy=TRUE)
sum_LabPr_RV_UV <- summary(SEM_LabPr_RV_UV)
sum_LabPr_RV_UV##
## Call:errorsarlm(formula = labpr_2017_RV_UV, listw = Russia.weights,
## zero.policy = TRUE)
##
## Residuals:
## Min 1Q Median 3Q Max
## -7.339217 -1.568437 -0.012222 1.383038 9.771580
##
## Type: error
## Coefficients: (asymptotic standard errors)
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) 134.0442462 18.1914227 7.3685 1.725e-13
## Rv_2017 -0.3015020 0.5140901 -0.5865 0.55756
## Uv_2017 0.5256910 0.6089462 0.8633 0.38798
## log(Inv_2017) -0.1557103 0.4814599 -0.3234 0.74638
## educ_2017 -0.0035936 0.0036895 -0.9740 0.33004
## Wage_2017 -0.3045703 0.1806852 -1.6856 0.09186
##
## Lambda: -0.25462, LR test value: 0.79569, p-value: 0.37239
## Asymptotic standard error: 0.27104
## z-value: -0.93943, p-value: 0.34751
## Wald statistic: 0.88253, p-value: 0.34751
##
## Log likelihood: -191.3756 for error model
## ML residual variance (sigma squared): 7.3993, (sigma: 2.7202)
## Number of observations: 79
## Number of parameters estimated: 8
## AIC: 398.75, (AIC for lm: 397.55)# sum_LabPr_RV_UV$Coef[, c(1,4)] %>% xtable()SEM_Emp_RV_UV <- errorsarlm(emp_2017_RV_UV, listw = Russia.weights, zero.policy=TRUE)
sum_Emp_RV_UV <- summary(SEM_Emp_RV_UV)
sum_Emp_RV_UV##
## Call:errorsarlm(formula = emp_2017_RV_UV, listw = Russia.weights,
## zero.policy = TRUE)
##
## Residuals:
## Min 1Q Median 3Q Max
## -0.0240285 -0.0047497 -0.0019357 0.0029796 0.0421506
##
## Type: error
## Coefficients: (asymptotic standard errors)
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -6.5370e-03 6.2903e-02 -0.1039 0.91723
## Rv_2017 -1.2833e-03 1.9811e-03 -0.6478 0.51713
## Uv_2017 3.5233e-03 2.0827e-03 1.6918 0.09069
## log(Inv_2017) 3.3945e-03 1.8907e-03 1.7954 0.07260
## educ_2017 -1.4115e-05 1.2933e-05 -1.0914 0.27509
## Wage_2017 -1.7146e-04 6.1562e-04 -0.2785 0.78061
##
## Lambda: 0.64862, LR test value: 18.228, p-value: 1.9602e-05
## Asymptotic standard error: 0.1151
## z-value: 5.6352, p-value: 1.7489e-08
## Wald statistic: 31.755, p-value: 1.7489e-08
##
## Log likelihood: 252.9047 for error model
## ML residual variance (sigma squared): 9.124e-05, (sigma: 0.009552)
## Number of observations: 79
## Number of parameters estimated: 8
## AIC: -489.81, (AIC for lm: -473.58)# sum_Emp_RV_UV$Coef[, c(1,4)] %>% xtable()SEM_Unemp_RV_UV <- errorsarlm(unemp_2017_RV_UV, listw = Russia.weights, zero.policy=TRUE)
sum_Unemp_RV_UV <- summary(SEM_Unemp_RV_UV)
sum_Unemp_RV_UV##
## Call:errorsarlm(formula = unemp_2017_RV_UV, listw = Russia.weights,
## zero.policy = TRUE)
##
## Residuals:
## Min 1Q Median 3Q Max
## -4.39172 -1.27483 -0.46471 0.43817 15.11471
##
## Type: error
## Coefficients: (asymptotic standard errors)
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) 54.0341714 16.1969696 3.3361 0.0008497
## Rv_2017 -0.8681169 0.5103454 -1.7010 0.0889359
## Uv_2017 -0.8698158 0.5337190 -1.6297 0.1031594
## log(Inv_2017) -1.2875275 0.4893523 -2.6311 0.0085113
## educ_2017 -0.0040946 0.0033228 -1.2323 0.2178434
## Wage_2017 -0.2915002 0.1579002 -1.8461 0.0648771
##
## Lambda: 0.73449, LR test value: 29.063, p-value: 7.006e-08
## Asymptotic standard error: 0.092918
## z-value: 7.9047, p-value: 2.6645e-15
## Wald statistic: 62.484, p-value: 2.6645e-15
##
## Log likelihood: -186.8151 for error model
## ML residual variance (sigma squared): 6.0784, (sigma: 2.4654)
## Number of observations: 79
## Number of parameters estimated: 8
## AIC: 389.63, (AIC for lm: 416.69)# sum_Unemp_RV_UV$Coef[, c(1,4)] %>% xtable()SAR, RV, UV
SAR_GRP_RV_UV <- lagsarlm(grp_2017_RV_UV, listw = Russia.weights, zero.policy=TRUE)
sum_GRP_RV_UV <- summary(SAR_GRP_RV_UV)
sum_GRP_RV_UV##
## Call:
## lagsarlm(formula = grp_2017_RV_UV, listw = Russia.weights, zero.policy = TRUE)
##
## Residuals:
## Min 1Q Median 3Q Max
## -1.601843 -0.381421 -0.020803 0.366851 1.859883
##
## Type: lag
## Coefficients: (asymptotic standard errors)
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -5.26362652 4.85216403 -1.0848 0.27801
## Rv_2017 0.22694118 0.13903419 1.6323 0.10262
## Uv_2017 0.76473134 0.15875648 4.8170 1.457e-06
## log(Inv_2017) 0.68249716 0.13290477 5.1352 2.818e-07
## educ_2017 0.00207507 0.00097229 2.1342 0.03282
## Wage_2017 0.02670350 0.04646894 0.5747 0.56553
##
## Rho: 0.41441, LR test value: 7.1133, p-value: 0.0076512
## Asymptotic standard error: 0.11595
## z-value: 3.5739, p-value: 0.00035165
## Wald statistic: 12.773, p-value: 0.00035165
##
## Log likelihood: -84.9298 for lag model
## ML residual variance (sigma squared): 0.49242, (sigma: 0.70172)
## Number of observations: 79
## Number of parameters estimated: 8
## AIC: 185.86, (AIC for lm: 190.97)
## LM test for residual autocorrelation
## test value: 17.003, p-value: 3.7317e-05# sum_GRP_RV_UV$Coef[, c(1,4)] %>% xtable()SAR_LabPr_RV_UV <- lagsarlm(labpr_2017_RV_UV, listw = Russia.weights, zero.policy=TRUE)
sum_LabPr_RV_UV <- summary(SAR_LabPr_RV_UV)
sum_LabPr_RV_UV##
## Call:lagsarlm(formula = labpr_2017_RV_UV, listw = Russia.weights,
## zero.policy = TRUE)
##
## Residuals:
## Min 1Q Median 3Q Max
## -7.391974 -1.605995 -0.067596 1.422568 9.812599
##
## Type: lag
## Coefficients: (asymptotic standard errors)
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) 154.9373027 32.1146394 4.8245 1.403e-06
## Rv_2017 -0.2186861 0.5364057 -0.4077 0.68350
## Uv_2017 0.5459102 0.6094575 0.8957 0.37040
## log(Inv_2017) -0.1805109 0.5027455 -0.3591 0.71956
## educ_2017 -0.0038268 0.0037172 -1.0295 0.30326
## Wage_2017 -0.2967812 0.1804183 -1.6450 0.09998
##
## Rho: -0.20944, LR test value: 0.65137, p-value: 0.41962
## Asymptotic standard error: 0.26232
## z-value: -0.79841, p-value: 0.42463
## Wald statistic: 0.63746, p-value: 0.42463
##
## Log likelihood: -191.4478 for lag model
## ML residual variance (sigma squared): 7.4261, (sigma: 2.7251)
## Number of observations: 79
## Number of parameters estimated: 8
## AIC: 398.9, (AIC for lm: 397.55)
## LM test for residual autocorrelation
## test value: 0.076501, p-value: 0.7821# sum_LabPr_RV_UV$Coef[, c(1,4)] %>% xtable()SAR_Emp_RV_UV <- lagsarlm(emp_2017_RV_UV, listw = Russia.weights, zero.policy=TRUE)
sum_Emp_RV_UV <- summary(SAR_Emp_RV_UV)
sum_Emp_RV_UV##
## Call:
## lagsarlm(formula = emp_2017_RV_UV, listw = Russia.weights, zero.policy = TRUE)
##
## Residuals:
## Min 1Q Median 3Q Max
## -0.0255106 -0.0042322 -0.0015043 0.0031514 0.0409784
##
## Type: lag
## Coefficients: (asymptotic standard errors)
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -1.7499e-02 6.2778e-02 -0.2787 0.78044
## Rv_2017 -7.2485e-04 1.8481e-03 -0.3922 0.69490
## Uv_2017 4.4786e-03 2.1079e-03 2.1247 0.03361
## log(Inv_2017) 3.8050e-03 1.7639e-03 2.1572 0.03099
## educ_2017 -1.6962e-05 1.2865e-05 -1.3185 0.18733
## Wage_2017 -2.4252e-04 6.2176e-04 -0.3901 0.69649
##
## Rho: 0.59543, LR test value: 21.882, p-value: 2.9001e-06
## Asymptotic standard error: 0.11764
## z-value: 5.0617, p-value: 4.156e-07
## Wald statistic: 25.621, p-value: 4.156e-07
##
## Log likelihood: 254.7317 for lag model
## ML residual variance (sigma squared): 8.8191e-05, (sigma: 0.009391)
## Number of observations: 79
## Number of parameters estimated: 8
## AIC: -493.46, (AIC for lm: -473.58)
## LM test for residual autocorrelation
## test value: 1.5874, p-value: 0.2077# sum_Emp_RV_UV$Coef[, c(1,4)] %>% xtable()SAR_Unemp_RV_UV <- lagsarlm(unemp_2017_RV_UV, listw = Russia.weights, zero.policy=TRUE)
sum_Unemp_RV_UV <- summary(SAR_Unemp_RV_UV)
sum_Unemp_RV_UV##
## Call:lagsarlm(formula = unemp_2017_RV_UV, listw = Russia.weights,
## zero.policy = TRUE)
##
## Residuals:
## Min 1Q Median 3Q Max
## -4.33587 -1.23364 -0.29277 0.67184 14.69214
##
## Type: lag
## Coefficients: (asymptotic standard errors)
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) 46.7447483 16.6151022 2.8134 0.004902
## Rv_2017 -0.3572664 0.4885261 -0.7313 0.464587
## Uv_2017 -0.8712627 0.5560522 -1.5669 0.117145
## log(Inv_2017) -1.4090245 0.4613569 -3.0541 0.002257
## educ_2017 -0.0061257 0.0033969 -1.8033 0.071338
## Wage_2017 -0.2516032 0.1637879 -1.5362 0.124501
##
## Rho: 0.71156, LR test value: 29.138, p-value: 6.7393e-08
## Asymptotic standard error: 0.089121
## z-value: 7.9843, p-value: 1.3323e-15
## Wald statistic: 63.748, p-value: 1.4433e-15
##
## Log likelihood: -186.7775 for lag model
## ML residual variance (sigma squared): 6.1189, (sigma: 2.4736)
## Number of observations: 79
## Number of parameters estimated: 8
## AIC: 389.55, (AIC for lm: 416.69)
## LM test for residual autocorrelation
## test value: 0.070019, p-value: 0.79131# sum_Unemp_RV_UV$Coef[, c(1,4)] %>% xtable()#Right tables, V ## SEM, V
SEM_GRP_V <- errorsarlm(grp_2017_V, listw = Russia.weights, zero.policy=TRUE)
sum_GRP_V <- summary(SEM_GRP_V)
sum_GRP_V##
## Call:
## errorsarlm(formula = grp_2017_V, listw = Russia.weights, zero.policy = TRUE)
##
## Residuals:
## Min 1Q Median 3Q Max
## -1.4572154 -0.3616371 0.0082359 0.3594635 1.6644839
##
## Type: error
## Coefficients: (asymptotic standard errors)
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -4.11947703 4.19264608 -0.9825 0.32583
## V_2017 0.59945416 0.08793841 6.8167 9.312e-12
## log(Inv_2017) 0.84441430 0.12672454 6.6634 2.676e-11
## educ_2017 0.00148510 0.00080593 1.8427 0.06537
## Wage_2017 0.05411408 0.04112019 1.3160 0.18817
##
## Lambda: 0.69705, LR test value: 22.768, p-value: 1.8275e-06
## Asymptotic standard error: 0.10281
## z-value: 6.7799, p-value: 1.203e-11
## Wald statistic: 45.966, p-value: 1.203e-11
##
## Log likelihood: -79.90729 for error model
## ML residual variance (sigma squared): 0.41077, (sigma: 0.64091)
## Number of observations: 79
## Number of parameters estimated: 7
## AIC: 173.81, (AIC for lm: 194.58)# sum_GRP_V$Coef[, c(1,4)] %>% xtable()SEM_LabPr_V <- errorsarlm(labpr_2017_V, listw = Russia.weights, zero.policy=TRUE)
sum_LabPr_V <- summary(SEM_LabPr_V)
sum_LabPr_V##
## Call:
## errorsarlm(formula = labpr_2017_V, listw = Russia.weights, zero.policy = TRUE)
##
## Residuals:
## Min 1Q Median 3Q Max
## -7.61870 -1.53586 0.10756 1.34730 10.05941
##
## Type: error
## Coefficients: (asymptotic standard errors)
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) 136.3692901 18.1140989 7.5284 5.129e-14
## V_2017 0.0841114 0.3303093 0.2546 0.79900
## log(Inv_2017) -0.1259115 0.4855179 -0.2593 0.79538
## educ_2017 -0.0024834 0.0034384 -0.7222 0.47014
## Wage_2017 -0.3255194 0.1798097 -1.8104 0.07024
##
## Lambda: -0.19552, LR test value: 0.50358, p-value: 0.47793
## Asymptotic standard error: 0.26435
## z-value: -0.73962, p-value: 0.45953
## Wald statistic: 0.54704, p-value: 0.45953
##
## Log likelihood: -191.76 for error model
## ML residual variance (sigma squared): 7.4886, (sigma: 2.7365)
## Number of observations: 79
## Number of parameters estimated: 7
## AIC: 397.52, (AIC for lm: 396.02)# sum_LabPr_V$Coef[, c(1,4)] %>% xtable()SEM_Emp_V <- errorsarlm(emp_2017_V, listw = Russia.weights, zero.policy=TRUE)
sum_Emp_V <- summary(SEM_Emp_V)
sum_Emp_V##
## Call:
## errorsarlm(formula = emp_2017_V, listw = Russia.weights, zero.policy = TRUE)
##
## Residuals:
## Min 1Q Median 3Q Max
## -0.0235151 -0.0047620 -0.0014592 0.0026884 0.0432237
##
## Type: error
## Coefficients: (asymptotic standard errors)
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) 3.7341e-03 6.3452e-02 0.0588 0.9531
## V_2017 1.0055e-03 1.3286e-03 0.7568 0.4492
## log(Inv_2017) 3.1601e-03 1.9132e-03 1.6517 0.0986
## educ_2017 -6.7154e-06 1.2208e-05 -0.5501 0.5823
## Wage_2017 -2.1739e-04 6.2334e-04 -0.3488 0.7273
##
## Lambda: 0.6554, LR test value: 18.827, p-value: 1.4313e-05
## Asymptotic standard error: 0.11342
## z-value: 5.7788, p-value: 7.5244e-09
## Wald statistic: 33.394, p-value: 7.5244e-09
##
## Log likelihood: 251.7265 for error model
## ML residual variance (sigma squared): 9.384e-05, (sigma: 0.0096871)
## Number of observations: 79
## Number of parameters estimated: 7
## AIC: -489.45, (AIC for lm: -472.63)# sum_Emp_V$Coef[, c(1,4)] %>% xtable()SEM_Unemp_V <- errorsarlm(unemp_2017_V, listw = Russia.weights, zero.policy=TRUE)
sum_Unemp_V <- summary(SEM_Unemp_V)
sum_Unemp_V##
## Call:
## errorsarlm(formula = unemp_2017_V, listw = Russia.weights, zero.policy = TRUE)
##
## Residuals:
## Min 1Q Median 3Q Max
## -4.39135 -1.27525 -0.46509 0.43801 15.11481
##
## Type: error
## Coefficients: (asymptotic standard errors)
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) 54.0308174 16.1122315 3.3534 0.0007982
## V_2017 -0.8689451 0.3382183 -2.5692 0.0101938
## log(Inv_2017) -1.2874281 0.4877799 -2.6394 0.0083062
## educ_2017 -0.0040971 0.0030934 -1.3245 0.1853521
## Wage_2017 -0.2914879 0.1577385 -1.8479 0.0646142
##
## Lambda: 0.73452, LR test value: 31.455, p-value: 2.0414e-08
## Asymptotic standard error: 0.092911
## z-value: 7.9056, p-value: 2.6645e-15
## Wald statistic: 62.498, p-value: 2.6645e-15
##
## Log likelihood: -186.8151 for error model
## ML residual variance (sigma squared): 6.0784, (sigma: 2.4654)
## Number of observations: 79
## Number of parameters estimated: 7
## AIC: 387.63, (AIC for lm: 417.08)# sum_Unemp_V$Coef[, c(1,4)] %>% xtable()SAR, V
SAR_GRP_V <- lagsarlm(grp_2017_V, listw = Russia.weights, zero.policy=TRUE)
sum_GRP_V <- summary(SAR_GRP_V)
sum_GRP_V##
## Call:lagsarlm(formula = grp_2017_V, listw = Russia.weights, zero.policy = TRUE)
##
## Residuals:
## Min 1Q Median 3Q Max
## -1.72443 -0.39031 -0.03862 0.35148 2.10420
##
## Type: lag
## Coefficients: (asymptotic standard errors)
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -4.43646678 4.94129824 -0.8978 0.369274
## V_2017 0.46980323 0.09440304 4.9766 6.472e-07
## log(Inv_2017) 0.70141368 0.13523092 5.1868 2.140e-07
## educ_2017 0.00288847 0.00092998 3.1060 0.001897
## Wage_2017 0.01695573 0.04759926 0.3562 0.721677
##
## Rho: 0.43781, LR test value: 7.6954, p-value: 0.0055362
## Asymptotic standard error: 0.11549
## z-value: 3.7908, p-value: 0.00015019
## Wald statistic: 14.37, p-value: 0.00015019
##
## Log likelihood: -87.4438 for lag model
## ML residual variance (sigma squared): 0.52332, (sigma: 0.72341)
## Number of observations: 79
## Number of parameters estimated: 7
## AIC: 188.89, (AIC for lm: 194.58)
## LM test for residual autocorrelation
## test value: 17.657, p-value: 2.6453e-05# sum_GRP_V$Coef[, c(1,4)] %>% xtable()SAR_LabPr_V <- lagsarlm(labpr_2017_V, listw = Russia.weights, zero.policy=TRUE)
sum_LabPr_V <- summary(SAR_LabPr_V)
sum_LabPr_V##
## Call:
## lagsarlm(formula = labpr_2017_V, listw = Russia.weights, zero.policy = TRUE)
##
## Residuals:
## Min 1Q Median 3Q Max
## -7.61160 -1.58671 0.13859 1.39787 10.06902
##
## Type: lag
## Coefficients: (asymptotic standard errors)
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) 153.0414023 31.8372152 4.8070 1.532e-06
## V_2017 0.1329236 0.3458757 0.3843 0.7007
## log(Inv_2017) -0.1649660 0.5037736 -0.3275 0.7433
## educ_2017 -0.0027275 0.0034636 -0.7875 0.4310
## Wage_2017 -0.3132298 0.1802191 -1.7380 0.0822
##
## Rho: -0.17144, LR test value: 0.45327, p-value: 0.50079
## Asymptotic standard error: 0.25959
## z-value: -0.66043, p-value: 0.50898
## Wald statistic: 0.43616, p-value: 0.50898
##
## Log likelihood: -191.7851 for lag model
## ML residual variance (sigma squared): 7.4991, (sigma: 2.7385)
## Number of observations: 79
## Number of parameters estimated: 7
## AIC: 397.57, (AIC for lm: 396.02)
## LM test for residual autocorrelation
## test value: 0.00032567, p-value: 0.9856# sum_LabPr_V$Coef[, c(1,4)] %>% xtable()SAR_Emp_V <- lagsarlm(emp_2017_V, listw = Russia.weights, zero.policy=TRUE)
sum_Emp_V <- summary(SAR_Emp_V)
sum_Emp_V##
## Call:lagsarlm(formula = emp_2017_V, listw = Russia.weights, zero.policy = TRUE)
##
## Residuals:
## Min 1Q Median 3Q Max
## -0.0251511 -0.0046484 -0.0017103 0.0029902 0.0422805
##
## Type: lag
## Coefficients: (asymptotic standard errors)
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -3.8653e-03 6.3309e-02 -0.0611 0.9513
## V_2017 1.5971e-03 1.2112e-03 1.3187 0.1873
## log(Inv_2017) 3.9399e-03 1.7733e-03 2.2218 0.0263
## educ_2017 -8.8339e-06 1.2106e-05 -0.7297 0.4656
## Wage_2017 -3.5976e-04 6.2815e-04 -0.5727 0.5668
##
## Rho: 0.60472, LR test value: 22.163, p-value: 2.505e-06
## Asymptotic standard error: 0.11786
## z-value: 5.1308, p-value: 2.8845e-07
## Wald statistic: 26.326, p-value: 2.8845e-07
##
## Log likelihood: 253.3943 for lag model
## ML residual variance (sigma squared): 9.1047e-05, (sigma: 0.0095419)
## Number of observations: 79
## Number of parameters estimated: 7
## AIC: -492.79, (AIC for lm: -472.63)
## LM test for residual autocorrelation
## test value: 1.6953, p-value: 0.19291# sum_Emp_V$Coef[, c(1,4)] %>% xtable()SAR_Unemp_V <- lagsarlm(unemp_2017_V, listw = Russia.weights, zero.policy=TRUE)
sum_Unemp_V <- summary(SAR_Unemp_V)
sum_Unemp_V##
## Call:
## lagsarlm(formula = unemp_2017_V, listw = Russia.weights, zero.policy = TRUE)
##
## Residuals:
## Min 1Q Median 3Q Max
## -4.22816 -1.25631 -0.29717 0.72377 14.79010
##
## Type: lag
## Coefficients: (asymptotic standard errors)
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) 45.6125959 16.5041991 2.7637 0.005715
## V_2017 -0.5942460 0.3170156 -1.8745 0.060861
## log(Inv_2017) -1.4192191 0.4594867 -3.0887 0.002010
## educ_2017 -0.0068755 0.0031531 -2.1806 0.029214
## Wage_2017 -0.2433058 0.1629150 -1.4935 0.135319
##
## Rho: 0.72246, LR test value: 31.159, p-value: 2.377e-08
## Asymptotic standard error: 0.086524
## z-value: 8.3498, p-value: < 2.22e-16
## Wald statistic: 69.72, p-value: < 2.22e-16
##
## Log likelihood: -186.9628 for lag model
## ML residual variance (sigma squared): 6.126, (sigma: 2.4751)
## Number of observations: 79
## Number of parameters estimated: 7
## AIC: 387.93, (AIC for lm: 417.08)
## LM test for residual autocorrelation
## test value: 0.22334, p-value: 0.63651# sum_Unemp_V$Coef[, c(1,4)] %>% xtable()#Common tables
stargazer(SEM_GRP_V, reg_labpr_2017_V, SAR_Emp_V, SAR_Unemp_V, SEM_GRP_RV_UV, reg_labpr_2017_RV_UV, SAR_Emp_RV_UV, SAR_Unemp_RV_UV, type = "text", omit = c('Constant'), keep.stat = c("n"), font.size = "tiny", title = 'Crossectional Spatial models, mixed types',
add.lines = list(c('model', 'SEM', 'OLS', 'SAR', 'SAR', 'SEM', 'OLS', 'SAR', 'SAR')), header=FALSE, dep.var.labels = NULL,
model.names = FALSE) #ci = T##
## Crossectional Spatial models, mixed types
## =========================================================================================================
## Dependent variable:
## -------------------------------------------------------------------------------------------
## log(GRP_2017) LabPr_2017 Emp_2017 Unemp_2017 log(GRP_2017) LabPr_2017 Emp_2017 Unemp_2017
## (1) (2) (3) (4) (5) (6) (7) (8)
## ---------------------------------------------------------------------------------------------------------
## V_2017 0.599*** 0.168 0.002 -0.594*
## (0.088) (0.359) (0.001) (0.317)
##
## Rv_2017 0.376*** -0.119 -0.001 -0.357
## (0.128) (0.561) (0.002) (0.489)
##
## Uv_2017 0.843*** 0.517 0.004** -0.871
## (0.135) (0.638) (0.002) (0.556)
##
## log(Inv_2017) 0.844*** -0.221 0.004** -1.419*** 0.867*** -0.244 0.004** -1.409***
## (0.127) (0.523) (0.002) (0.459) (0.123) (0.526) (0.002) (0.461)
##
## educ_2017 0.001* -0.003 -0.00001 -0.007** 0.001 -0.004 -0.00002 -0.006*
## (0.001) (0.004) (0.00001) (0.003) (0.001) (0.004) (0.00001) (0.003)
##
## Wage_2017 0.054 -0.312* -0.0004 -0.243 0.059 -0.298 -0.0002 -0.252
## (0.041) (0.187) (0.001) (0.163) (0.040) (0.189) (0.001) (0.164)
##
## ---------------------------------------------------------------------------------------------------------
## model SEM OLS SAR SAR SEM OLS SAR SAR
## Observations 79 79 79 79 79 79 79 79
## =========================================================================================================
## Note: *p<0.1; **p<0.05; ***p<0.01stargazer(SEM_GRP_V, SEM_LabPr_V, SEM_Emp_V, SEM_Unemp_V, SEM_GRP_RV_UV, SEM_LabPr_RV_UV, SEM_Emp_RV_UV, SEM_Unemp_RV_UV, type = "text", omit = c('Constant'), keep.stat = c("n"), font.size = "tiny", title = 'Crossectional Spatial models, Spatial Error model',
add.lines = list(c('model', 'SEM', 'SEM', 'SEM', 'SEM', 'SEM', 'SEM', 'SEM', 'SEM')), header=FALSE, dep.var.labels = NULL,
model.names = FALSE)##
## Crossectional Spatial models, Spatial Error model
## =========================================================================================================
## Dependent variable:
## -------------------------------------------------------------------------------------------
## log(GRP_2017) LabPr_2017 Emp_2017 Unemp_2017 log(GRP_2017) LabPr_2017 Emp_2017 Unemp_2017
## (1) (2) (3) (4) (5) (6) (7) (8)
## ---------------------------------------------------------------------------------------------------------
## V_2017 0.599*** 0.084 0.001 -0.869**
## (0.088) (0.330) (0.001) (0.338)
##
## Rv_2017 0.376*** -0.302 -0.001 -0.868*
## (0.128) (0.514) (0.002) (0.510)
##
## Uv_2017 0.843*** 0.526 0.004* -0.870
## (0.135) (0.609) (0.002) (0.534)
##
## log(Inv_2017) 0.844*** -0.126 0.003* -1.287*** 0.867*** -0.156 0.003* -1.288***
## (0.127) (0.486) (0.002) (0.488) (0.123) (0.481) (0.002) (0.489)
##
## educ_2017 0.001* -0.002 -0.00001 -0.004 0.001 -0.004 -0.00001 -0.004
## (0.001) (0.003) (0.00001) (0.003) (0.001) (0.004) (0.00001) (0.003)
##
## Wage_2017 0.054 -0.326* -0.0002 -0.291* 0.059 -0.305* -0.0002 -0.292*
## (0.041) (0.180) (0.001) (0.158) (0.040) (0.181) (0.001) (0.158)
##
## ---------------------------------------------------------------------------------------------------------
## model SEM SEM SEM SEM SEM SEM SEM SEM
## Observations 79 79 79 79 79 79 79 79
## =========================================================================================================
## Note: *p<0.1; **p<0.05; ***p<0.01stargazer( SAR_GRP_V, SAR_LabPr_V, SAR_Emp_V, SAR_Unemp_V, SAR_GRP_RV_UV, SAR_LabPr_RV_UV, SAR_Emp_RV_UV, SAR_Unemp_RV_UV, type = "text", omit = c('Constant'), keep.stat = c("n"), font.size = "tiny", title = 'Crossectional Spatial models, Spatial Autoregressive model',
add.lines = list(c('model', 'SAR', 'SAR', 'SAR', 'SAR', 'SAR', 'SAR', 'SAR', 'SAR')), header=FALSE, dep.var.labels = NULL,
model.names = FALSE)##
## Crossectional Spatial models, Spatial Autoregressive model
## =========================================================================================================
## Dependent variable:
## -------------------------------------------------------------------------------------------
## log(GRP_2017) LabPr_2017 Emp_2017 Unemp_2017 log(GRP_2017) LabPr_2017 Emp_2017 Unemp_2017
## (1) (2) (3) (4) (5) (6) (7) (8)
## ---------------------------------------------------------------------------------------------------------
## V_2017 0.470*** 0.133 0.002 -0.594*
## (0.094) (0.346) (0.001) (0.317)
##
## Rv_2017 0.227 -0.219 -0.001 -0.357
## (0.139) (0.536) (0.002) (0.489)
##
## Uv_2017 0.765*** 0.546 0.004** -0.871
## (0.159) (0.609) (0.002) (0.556)
##
## log(Inv_2017) 0.701*** -0.165 0.004** -1.419*** 0.682*** -0.181 0.004** -1.409***
## (0.135) (0.504) (0.002) (0.459) (0.133) (0.503) (0.002) (0.461)
##
## educ_2017 0.003*** -0.003 -0.00001 -0.007** 0.002** -0.004 -0.00002 -0.006*
## (0.001) (0.003) (0.00001) (0.003) (0.001) (0.004) (0.00001) (0.003)
##
## Wage_2017 0.017 -0.313* -0.0004 -0.243 0.027 -0.297* -0.0002 -0.252
## (0.048) (0.180) (0.001) (0.163) (0.046) (0.180) (0.001) (0.164)
##
## ---------------------------------------------------------------------------------------------------------
## model SAR SAR SAR SAR SAR SAR SAR SAR
## Observations 79 79 79 79 79 79 79 79
## =========================================================================================================
## Note: *p<0.1; **p<0.05; ***p<0.01