Question: why negative binominal part of hurdle model does not provide coefficients for intercept and word count?
I have counted positive emotional words (Y) in some conversations that have a different length. Positive emotion variable has a Poission distribution. A Poission regression controlling for the total amount of words in the conversation do not satisfy a dispersion test. (Test suggest that the data overdispersed). I choose to run a hurdle model instead. I am trying to fit the data in a hurdle model figuring out whether the presence and amount of positive words (Y) depend on the advice given in the conversation (X). I also control for the total word count in the conversations (WC) since conversations have a different length. Here is what I have:
Y <- cbind(PosEmotCount) X <- cbind(AdviceAS, WC) hnegbin <- hurdle(Y ~ X, link = "logit", dist = "negbin") Pearson residuals: Min 1Q Median 3Q Max -1.7208 -0.7297 -0.2498 0.4924 3.5845 Count model coefficients (truncated negbin with log link): Estimate Std. Error z value Pr(>|z|) (Intercept) 0.7551942 NA NA NA XAdviceAS -0.0541452 0.1096667 -0.494 0.621 XWC 0.0003255 NA NA NA Log(theta) 1.3652328 NA NA NA Zero hurdle model coefficients (binomial with logit link): Estimate Std. Error z value Pr(>|z|) (Intercept) 0.3011979 0.3781910 0.796 0.426 XAdviceAS 0.7367846 0.6598110 1.117 0.264 XWC 0.0009060 0.0001132 8.002 1.22e-15 *** --- Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
QUESTION: why negative binominal part of the hurdle model does not provide coefficients for intercept and word count?