Visualize exit variable

## [1] "Visualize distribution of outcome variables"

Testing TWFE

# note that this is an early pass; no  diagnostics, etc. (like, for example, it seems that there is serial correlation in this model -- meaning that a lag term may need to be included?)

# make outcome variable
dfModel$exitRate = dfModel$exited/dfModel$child

# run model
twfe_exitRt <- plm(exitRate ~ unempRt + singPar + foodIns + chpovRt + LessHS + gspPc + policy_updated + crimeRt + pctBlk + tanfBen_3fam + snapPpl + mediPpl + tanfPpl + welfarePc + tanfPpl*welfarePc + incarRt + deathRt + policePc + factor(year), data=dfModel, effect="twoway", model="within", index="state")

summary(twfe_exitRt)
## Twoways effects Within Model
## 
## Call:
## plm(formula = exitRate ~ unempRt + singPar + foodIns + chpovRt + 
##     LessHS + gspPc + policy_updated + crimeRt + pctBlk + tanfBen_3fam + 
##     snapPpl + mediPpl + tanfPpl + welfarePc + tanfPpl * welfarePc + 
##     incarRt + deathRt + policePc + factor(year), data = dfModel, 
##     effect = "twoway", model = "within", index = "state")
## 
## Unbalanced Panel: n = 50, T = 18-19, N = 946
## 
## Residuals:
##        Min.     1st Qu.      Median     3rd Qu.        Max. 
## -3.7307e-03 -4.1697e-04  1.1019e-05  3.5086e-04  4.3683e-03 
## 
## Coefficients:
##                      Estimate  Std. Error t-value  Pr(>|t|)    
## unempRt            1.5462e-03  3.8291e-03  0.4038  0.686471    
## singPar           -1.8364e-03  2.0317e-03 -0.9038  0.366335    
## foodIns            2.6057e-03  1.3372e-03  1.9487  0.051657 .  
## chpovRt            1.6817e-03  2.2342e-03  0.7527  0.451827    
## LessHS            -9.3926e-03  3.7804e-03 -2.4845  0.013161 *  
## gspPc             -2.2580e-08  9.2026e-09 -2.4536  0.014339 *  
## policy_updated     5.4424e-04  1.2460e-04  4.3677 1.408e-05 ***
## crimeRt            2.1375e-05  9.2912e-06  2.3006  0.021653 *  
## pctBlk            -1.0170e-02  6.1416e-03 -1.6559  0.098104 .  
## tanfBen_3fam      -2.1939e-06  6.8765e-07 -3.1904  0.001472 ** 
## snapPpl           -1.5504e-03  3.3936e-04 -4.5685 5.631e-06 ***
## mediPpl            7.2385e-04  1.6026e-04  4.5169 7.152e-06 ***
## tanfPpl           -1.4158e-03  7.1193e-04 -1.9886  0.047060 *  
## welfarePc         -2.2784e-05  2.0289e-04 -0.1123  0.910614    
## incarRt            1.6067e-04  5.5702e-05  2.8844  0.004020 ** 
## deathRt            2.1384e-04  1.5350e-04  1.3931  0.163939    
## policePc           2.4923e-04  1.6286e-04  1.5304  0.126288    
## tanfPpl:welfarePc  8.5832e-04  3.6117e-04  2.3765  0.017696 *  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Total Sum of Squares:    0.00059939
## Residual Sum of Squares: 0.00048281
## R-Squared:      0.1945
## Adj. R-Squared: 0.11489
## F-statistic: 11.5365 on 18 and 860 DF, p-value: < 2.22e-16

Glmer Model: Rate of exit

exitModel <- glmer(exited ~ scale(unempRt) + scale(singPar) + scale(foodIns) + scale(chpovRt) + scale(LessHS) + scale(gspPc) + scale(policy_updated) + scale(crimeRt) + scale(pctBlk) + scale(tanfBen_3fam) + scale(snapPpl) + scale(mediPpl) + scale(tanfPpl) + scale(welfarePc) + scale(tanfPpl)*scale(welfarePc) + scale(incarRt) + scale(deathRt) + scale(policePc) + (1 | state) + (1 | year) + (1 | obsN), offset=log(child),data=dfModel,family=poisson)

summary(exitModel)
## Generalized linear mixed model fit by maximum likelihood (Laplace
##   Approximation) [glmerMod]
##  Family: poisson  ( log )
## Formula: 
## exited ~ scale(unempRt) + scale(singPar) + scale(foodIns) + scale(chpovRt) +  
##     scale(LessHS) + scale(gspPc) + scale(policy_updated) + scale(crimeRt) +  
##     scale(pctBlk) + scale(tanfBen_3fam) + scale(snapPpl) + scale(mediPpl) +  
##     scale(tanfPpl) + scale(welfarePc) + scale(tanfPpl) * scale(welfarePc) +  
##     scale(incarRt) + scale(deathRt) + scale(policePc) + (1 |  
##     state) + (1 | year) + (1 | obsN)
##    Data: dfModel
##  Offset: log(child)
## 
##      AIC      BIC   logLik deviance df.resid 
##  14869.0  14975.7  -7412.5  14825.0      924 
## 
## Scaled residuals: 
##      Min       1Q   Median       3Q      Max 
## -0.73703 -0.06391  0.00341  0.06037  0.77690 
## 
## Random effects:
##  Groups Name        Variance Std.Dev.
##  obsN   (Intercept) 0.024198 0.15556 
##  state  (Intercept) 0.112824 0.33589 
##  year   (Intercept) 0.001755 0.04189 
## Number of obs: 946, groups:  obsN, 946; state, 50; year, 19
## 
## Fixed effects:
##                                  Estimate Std. Error  z value Pr(>|z|)    
## (Intercept)                     -5.570617   0.048817 -114.113  < 2e-16 ***
## scale(unempRt)                  -0.003852   0.014083   -0.274  0.78447    
## scale(singPar)                  -0.006352   0.031100   -0.204  0.83817    
## scale(foodIns)                   0.011805   0.009466    1.247  0.21235    
## scale(chpovRt)                   0.005679   0.021498    0.264  0.79166    
## scale(LessHS)                   -0.067051   0.021923   -3.058  0.00223 ** 
## scale(gspPc)                    -0.027183   0.018837   -1.443  0.14900    
## scale(policy_updated)            0.198845   0.030306    6.561 5.34e-11 ***
## scale(crimeRt)                   0.018862   0.016323    1.156  0.24787    
## scale(pctBlk)                   -0.242922   0.049626   -4.895 9.83e-07 ***
## scale(tanfBen_3fam)             -0.040660   0.022584   -1.800  0.07180 .  
## scale(snapPpl)                  -0.085148   0.013998   -6.083 1.18e-09 ***
## scale(mediPpl)                   0.057520   0.014348    4.009 6.10e-05 ***
## scale(tanfPpl)                   0.008092   0.010808    0.749  0.45403    
## scale(welfarePc)                 0.066627   0.017045    3.909 9.27e-05 ***
## scale(incarRt)                   0.060365   0.021333    2.830  0.00466 ** 
## scale(deathRt)                   0.009264   0.007356    1.259  0.20793    
## scale(policePc)                  0.050091   0.015572    3.217  0.00130 ** 
## scale(tanfPpl):scale(welfarePc)  0.015641   0.008289    1.887  0.05917 .  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Correlation matrix not shown by default, as p = 19 > 12.
## Use print(x, correlation=TRUE)  or
##     vcov(x)        if you need it

Glmer R2

## [1] "Conditional R2 (fixed + random effects) is fairly small:"
## Warning: mu of 0.0 is too close to zero, estimate of random effect variances may
##   be unreliable.
## # R2 for Mixed Models
## 
##   Conditional R2: 0.034
##      Marginal R2: 0.014

Glmer ICC

## [1] "~2% of variance is due to differences between groups:"
## Warning: mu of 0.0 is too close to zero, estimate of random effect variances may
##   be unreliable.
## # Intraclass Correlation Coefficient
## 
##     Adjusted ICC: 0.020
##   Unadjusted ICC: 0.020

Glmer Diagnostics

Multicollinearity

Singularity & overdispersion

## [1] "no singularity detected:"
## [1] 0.04189393
## # Overdispersion test
## 
##        dispersion ratio =  0.019
##   Pearson's Chi-Squared = 17.736
##                 p-value =      1
## No overdispersion detected.

Plot random effects

## $obsN

## 
## $state

## 
## $year

Residuals (less reliable for glmer?)