pR2 function fails on fully populated R object

[Joseph White]

Hi,

The R package, pscl, is meant for statistical analysis. The pR2 function that calculates "pseudo R-sq" has been failing with error:

"no applicable method for 'pR2' applied to an object of class "NULL".

I know this means that R thinks data object sent to pR2 was null, but I have proof that the glm package that created the data object produced the expected output data objects. That is, the following code works up to the call to pR2.

*************************************
glm_modeling <- function(formula=NULL, dataSets=NULL){
fit_glm <- glm(formula, dataSets, family=binomial(),control=list(maxit=100))
message(str(fit_glm))
glm_summary <- summary(fit_glm)
list(fit_glm, glm_summary)
}

[snip]

warning_catch <- tryCatch(
{fit_list <- glm_modeling(formula=formula1, dataSets=reg_table)
message(dim(fit_list))
message(length(fit_list))
},
error=function (e) e,
warning=function(w) w
)

[snip]

glm_fit = fit_list[[1]]
glm_summary = fit_list[[2]]
pv_coef <- glm_summary$coefficients

# This call fails with error: " no applicable method for 'pR2' applied to an object of class "NULL" "
GOF <- as.data.frame(pR2(glm_fit))
*************************************

I know that the glm_fit object is populated because I checked it with 'message'.

So, has anyone else seen this behavior from the pR2 function in the pscl library? The installed pscl library is version pscl_1.4.9. Is there a later version? Any idea why pR2 would fail when given a fully populated object?

Any help is appreciated.

JoeW