Currently qgam calls na.omit( data ), but this is a serious problem because data might contain a column full of NAs, which is not included in the fit. Hence excluding the corresponding rows does not make sense.
This commit 3bee94a15b328c781cd4cbd9217b0301fe95f26b should have fixed the problem, we now select only the relevant columns of data, and then we drop out the rows containing NAs.
Currently qgam calls na.omit( data ), but this is a serious problem because data might contain a column full of NAs, which is not included in the fit. Hence excluding the corresponding rows does not make sense.