I struggle with modeling a no-choice option within the mlogit package to estimate part worths from conjoint data. I have choice data from 600 respondents (respid). Each respondent choose between 3 hypothetical products (alt: A-C) and a no choice option (alt: D). Everyone does the choice 4 times (ques).
I prepare the data the following:
data.test<-mlogit.data(data=robottest, choice="choice", shape="long", varying = 4:7, alt.var = "alt", alt.levels = paste("pos",1:4),id.var = "respid")
I get the following error message but it does create the data anyway.
Warning messages:
1: In mlogit.data(data = robottest, choice = "choice", shape = "long", :
variable alt exists and will be replaced
2: Setting row names on a tibble is deprecated.
I than want to estimate the model with the following code
m1<-mlogit(data=data.test, choice~apperance+features+brand+price, nests = list(Buy=c("A","B","C"), NoBuy=c("D")), unscaled = TRUE)
That does not work and I get the error saying
Error in solve.default(H, g[!fixed]) :
system is computationally singular: reciprocal condition number = 4.91778e-24
Do somebody know how to solve this issue? Would you model the no-choice option in general the same as I did? Any help is more than welcommed! Thank you very much.
Bests, Michael
