Dera
2015-Jul-27 22:01 UTC
[R] Error “system is exactly singular: U[6,6] = 0” when doing nested logit using mlogit package in R
have a problem about doing nested logit using mlogit package in R. We did a survey about asking people to choose from different insurance policies. To be specific, respondents were asked ?Would you buy Policy A, Policy B, or neither?" (Each choice is defined by the deductible and premium associated with them) Here is a brief example of our data named "mydata": RespondentID Choice alt Deductible Premium 1 False PolicyA 500 1000 1 TRUE PolicyB 1000 500 1 False Neither 0 0 2 TRUE PolicyA 250 5000 2 False PolicyB 500 500 2 False Neither 0 0 3 TRUE PolicyA 500 2000 3 False PolicyB 5000 1000 3 False Neither 0 0 ... I'd like to use nested logit here, because "Policy A" and "Policy B" are much similar than "Neither". So I tried to use the code to fit a nested logit model: nl.1 <- mlogit (Choice~Deductible+Premium, mydata,shape='long', choice 'Choice', nests=list(buy=c('PolicyA','PolicyA'), no='Neither'), alt.var "alt") But it always gave the error warning: Error in solve.default(H, g[!fixed]) : Lapack routine dgesv: system is exactly singular: U[6,6] = 0 If I do like this: nl.2 <- mlogit (Choice~Deductible+Premium, mydata,shape='long', choice 'Choice', nests=list(buy=c('PolicyA','PolicyA'), no='Neither'), alt.var "alt", un.nest.el = TRUE) It works. But I don't think the two nests I mention above have unique elasticity, what "un.nest.el = TRUE" means. Does anyone know about this? How can I avoid the error in the first code? Thank you! -- View this message in context: http://r.789695.n4.nabble.com/Error-system-is-exactly-singular-U-6-6-0-when-doing-nested-logit-using-mlogit-package-in-R-tp4710437.html Sent from the R help mailing list archive at Nabble.com.