I'm currently working on a project where my model is a GLM with a beta distribution. The dependent variable is bimodal. I've done this before, in Stata, but I'm having some difficulty interpreting what R has put out. I ran the model using the VGLM command from the VGAM package. For some reason, it has produced two coefficient for each variable. This is the part that confuses me--I'm not sure how to interpret this, and I haven't been able to find any help so far. My key variable of interest is logoil_gas_valuePOP_1. Any help here would be extremely appreciated--consulting my stats teacher produced nothing.
Pearson Residuals:
Min 1Q Median 3Q Max
elogit(mu, min = 0, max = 12) -6.4754 -0.34561 0.048202 0.42250 6.32340
log(phi) -23.6802 0.32592 0.592333 0.68735 0.73015
Coefficients:
Estimate Std. Error z value
(Intercept):1 -0.2731986 / 0.0513944 / -5.315728
(Intercept):2 0.9116027 / 0.1266222 / 7.199393
polity_5:1 0.1267469 / 0.0016007 / 79.180001
polity_5:2 0.0481726 / 0.0037464 / 12.858440
logGDP_cap2000_sup_1:1 0.0945281 / 0.0067247 / 14.056858
logGDP_cap2000_sup_1:2 0.3542733 / 0.0170221 / 20.812500
logoil_gas_valuePOP_1:1 -0.0330833 / 0.0035020 / -9.447096
logoil_gas_valuePOP_1:2 -0.0082593 / 0.0094502 / -0.873979
Number of linear predictors: 2
Names of linear predictors: elogit(mu, min = 0, max = 12), log(phi)
Dispersion Parameter for betaff family: 1
Log-likelihood: -8592.995 on 10366 degrees of freedom
Number of iterations: 25
Here is the code:
library("foreign")
rossdatahw <- read.dta("c:/Users/Amanda/Desktop/PS 531/Ross Replication/rossdata1.dta")
library(mosaic)
library("VGAM")
rossdatahw$polity01 <- with(rossdatahw,polity/10)
cor(rossdatahw[,c("polity","polity01")], use="complete.obs")
rossdatahw$polity1to11 <- rossdatahw$polity+1
bootrossyears <- resample(rossdatahw, within=rossdatahw$year, size=1000, replace=TRUE)
bdata <- bootrossyears
fit1 <- vglm(polity1to11~polity_5 + logGDP_cap2000_sup_1 +
logoil_gas_valuePOP_1 + latin + latin_logoil_1 + latin +
dem_prior_update + gdpgrowth_alt_new_1 + mus_pct_sup +
YEAR1960_64 + YEAR1965_69 + YEAR1970_74 + YEAR1975_79 +
YEAR1980_84 + YEAR1985_89 + YEAR1990_94 + YEAR1995_99,
data=rossdatahw[!is.na(rossdatahw$polity01),],
na.action=na.omit,
family=betaff(A=0,B=12,lmu = elogit(min = 0, max = 12)),
##family=beta.ab(A=0, B=12),
trace=TRUE)
coef(fit1, matrix=TRUE)
coef(fit1)
summary(fit1)
