It is rather an opinion than an issue. I was trying to predict caste~age+sex for a living standard survey data. It happened that some of the PSUs have a single caste group. As a result, I got a rather abstruse error from the simRelation function.
Error in multinom(caste ~ age + sex, weights = weights, data = dataSampleWork, : need two or more classes to fit a multinom model
(In hindsight, it is perfectly reasonable and explains the problem quite well)
It happened that while the dataset had a lot of variance in caste, some PSUs were homogenous. It would be helpful if there was a warning of some sort. Like inserting something like this in simulateValues function after line number 37: dataSampleWork <- dataSample[indSampleHead, ]
if(length(unique(dataSampleWork[, eval(parse(text = cur.var))]))==1){cat("Error: single level in",x,"", curStrata, "of sample")}
Also, @bernhard-da is it a good idea to fill in values from the sample in population if the stratum has a homogenous value. Would it be possible to implement inside this package? In my case using the homogenous value of the caste to fill all households in that PSU.
It is rather an opinion than an issue. I was trying to predict caste~age+sex for a living standard survey data. It happened that some of the PSUs have a single caste group. As a result, I got a rather abstruse error from the simRelation function. Error in multinom(caste ~ age + sex, weights = weights, data = dataSampleWork, : need two or more classes to fit a multinom model
(In hindsight, it is perfectly reasonable and explains the problem quite well) It happened that while the dataset had a lot of variance in caste, some PSUs were homogenous. It would be helpful if there was a warning of some sort. Like inserting something like this in simulateValues function after line number 37: dataSampleWork <- dataSample[indSampleHead, ]
if(length(unique(dataSampleWork[, eval(parse(text = cur.var))]))==1){cat("Error: single level in",x,"", curStrata, "of sample")}
Also, @bernhard-da is it a good idea to fill in values from the sample in population if the stratum has a homogenous value. Would it be possible to implement inside this package? In my case using the homogenous value of the caste to fill all households in that PSU.