manuelreif / svypvpack Goto Github PK

Levels of a factor describing proficiency levels in a data frame returned by svyPVlevel() has levels sorted alphabetically. See:

data(svy_example1)
erg_l <- svyPVlevel(by = ~ sex,
                    svydat=svy.exrep, 
                    pvs=c("plaus1","plaus2","plaus3"),
                    CATDEF=c(0,310,322,400),
                    levlab = c("c", "b", "a"),
                    colN = TRUE)
levels(erg_l$levels)

It will be better if this levels were ordered in the same way as they appear levlab argument to svyPVlevel().

Columns being named Proportion and SE.Proportion if there are many groups defined by by are named Percent and SE.Percent if there is only one group. See:

data(svy_example1)
svyPVpm(by = ~sex,
        svydat = svy.exrep,
        pvs = c("plaus1","plaus2","plaus3"))
svyPVpm(by = ~sex,
        svydat = subset(svy.exrep, sex == "m"),
        pvs = c("plaus1","plaus2","plaus3"))

This package is very useful however it will be even better if it can be used to analyze also variables that aren't PVs. Simply performing analysis on large scale assessment data typically involves dealing with regular variables along with dealing with PVs and it would be very convenient to have one tool that you can use for both of these (with most important thing is getting results in the same format).
Now with only one variable specified with pvs argument to svyPVpack function point estimates are obtained but standard errors estimates are set to NAs. I assume that's because measurement error can't be computed in such a case so its value is set to NA and this propagates while summing with sampling error.

Can you modify package in such a way, that with only one variable specified with pvs argument standard errors being computed taking into account solely sampling error? (perhaps with some message printed about such a behavior)