compareML not providing p-values

Question

I have two nested gamms and am trying to use compareML to compare the models because they have similar AIC values but gam10 has a lower BIC. The function runs without an error message, but no p-value is reported even with specifying that I would like the p-value reported.

Two gamms:

   gam8<-bam(Numberpertow ~ Stratum+ClosArea+s(interval, by=CruiseID) +s(StationID, bs = 're'),data=l.data,method = "fREML",family=nb())

gam10<-bam(Numberpertow ~ Stratum+s(interval)+ClosArea+s(StationID, bs = 're'),data=l.data,method = "fREML",family=nb())

#compareML code

compareML(gam10, gam8, signif.stars=T, suggest.report = T,
          print.output = TRUE)

function output:

gam10: Numberpertow ~ Stratum + s(interval, k = 6) + ClosArea + 
    s(StationID, bs = "re")

gam8: Numberpertow ~ Stratum + ClosArea + s(interval, by = CruiseID) + CruiseID + s(StationID, bs = "re")

Model gam10 preferred: lower fREML score (115.035), and lower df (6.000).
-----
  Model    Score Edf Difference     Df
1  gam8 38899.94  32                  
2 gam10 38784.90  26    115.035 -6.000

AIC difference: 204.73, model gam8 has lower AIC.

I have tried using ML for the smooth parameter estimation method and get the same result. The compareML help file indicates "The order of the two models is not important. Model comparison is only implemented for the methods GCV, fREML, REML, and ML."

The output indicates gam10 is preferred, but reviewers are looking for a p-value.

I am unsure how to provide a reproduceable example based on the size of my original data and not wanting to take a small subset because results may be different.

compareML not providing p-values

0 Answers