Error in MICE imputation using quickpred - system is computationally singular

[studentaccount221]

I'm trying to impute missing values in my dataset using MICE. I have a dataset consisting of 116 obs. of 134 variables. My dataset contains numerical, categorical and binary data. All variables with more than 50% missing data have been dropped. I used the quickpred function to select the predictors. However, I keep getting an error:

Error in solve.default(xtx + diag(pen)):
system is computationally singular: reciprocal condition number = 1.77279e-16

This is the code I use

inlist <- c("r_age", "r_sex")
exlist <- c("studysubjectid", "studyid") 
pred_matrix <- quickpred(dataset, minpuc = 0.5, include = inlist, exclude = exlist)
imputed_data <- mice(dataset, m = 20, maxit = 20, pred = pred_matrix, printFlag = FALSE, seed=1234)

I have tried using different values for mincor and minpuc, however the results do not change. I want to know what causes this error (multicollinearity?) and how to solve it?

[stefvanbuuren]

Do you have a categorical variable in your data with many many response categories, for example, an ID variable?

[studentaccount221]

Yes I do, but I left this out using exclude = exlist <- c("studysubjectid", "studyid") in quickpred

[stefvanbuuren]

I would advise you to reduce the number of variables based on substantive grounds. Keep what is important for answering your research question, and compress/ditch the rest.