The goal of hglmbc2 is to make inferences in Small Area Estimation based on Hierarchical (h − )likelihood approach with bias correction. The model parameters are obtained through an iterative approximation based on Newton Raphson method combined with bias correction of estimates. The bias correction approach enhances the accuracy of maximum hierarchical likelihood estimates (MHLEs). This R package can be used to obtain improved MHLEs for fixed effects, random effects, and dispersion parameters for exponential family distributions with random effect u ∼ N(0, σ2).
Installation
You can install the released version of hglmbc2 from CRAN with:
install.packages("hglmbc2")
or, you can install the development version of hglmbc2 using devtools with:
# devtools::install_github("niroshar/hglmbc2", force = TRUE)Example: method 1
Binomial-Normal HGLM is also known as the mixed logit model in GLM family with the binary response variable and the random effect u ∼ N(0, σ2).
library(hglmbc2) # basic example code data <- eversmoke mformula <- "smoke_ever ~ as.factor(age) + as.factor(gender) + as.factor(race) + as.factor(year) + povt_rate" dom <- "county" y.family <- "binomial" rand.family <- "gaussian" # Fit the model hglmbc.fit <- hglmbc(data = eversmoke, mformula, dom = "county", y.family = "binomial") #> Fixed effects and random effects are not defined, selected by variable type !!!! ## MHLEs of fixed effects hglmbc.fit$est.fe #> Estimate Std.Error Z Value P(>|Z|) #> Intercept -1.77212 0.15260 -11.61284 0.00000 #> age2 0.47708 0.05213 9.15174 0.00000 #> age3 1.00272 0.06069 16.52200 0.00000 #> gender2 0.13512 0.02828 4.77793 0.00000 #> povtrate 3.80385 0.86779 4.38338 0.00001 #> race2 -0.47891 0.05268 -9.09093 0.00000 #> race3 0.04269 0.03866 1.10424 0.26949 #> race4 -0.17499 0.04980 -3.51386 0.00044 #> year2 -0.32813 0.05561 -5.90056 0.00000 # The distribution of y|u not defined, hglmbc.fit1 <- hglmbc(data = eversmoke, mformula, dom = "county") #> Fixed effects and random effects are not defined, selected by variable type !!!! # Model fit summary hglmbc.fit1$summary #> $`Model formula` #> [1] "smoke_ever ~ as.factor(age) + as.factor(gender) + as.factor(race) + as.factor(year) + povt_rate" #> #> $`random effect` #> [1] "county" #> #> $`fixed effects estimates` #> Estimate Std.Error Z Value P(>|Z|) #> Intercept -1.77212 0.15260 -11.61284 0.00000 #> age2 0.47708 0.05213 9.15174 0.00000 #> age3 1.00272 0.06069 16.52200 0.00000 #> gender2 0.13512 0.02828 4.77793 0.00000 #> povtrate 3.80385 0.86778 4.38343 0.00001 #> race2 -0.47891 0.05268 -9.09093 0.00000 #> race3 0.04269 0.03866 1.10424 0.26949 #> race4 -0.17499 0.04980 -3.51386 0.00044 #> year2 -0.32813 0.05561 -5.90056 0.00000 #> #> $`dispersion paramerer` #> [,1] #> [1,] 0.20757 #> #> $`hglm model inference` #> AIC BIC hLik #> 1 29221.83 29303.13 -14600.91 #> #> $iter #> [1] 11 #> #> $` ` #> [1] "Converged in 11 iterations with tol = 5.30948712151869e-06." #> #> attr(,"class") #> [1] "summary.hglm.fit"
Example: method 2
# mformula is not defined, data <- eversmoke resp <- "smoke_ever" dom <- "county" fe.disc <- c("year","gender","race","age") fe.cont <- "povt_rate" # hglmbc.fit <- hglmbc(data = eversmoke, resp = "smoke_ever", dom = "county",fe.disc = fe.disc,fe.cont = fe.cont, y.family = "binomial") # hglmbc.fit
