# Predicting logistic regression in R with missing values

Can logistic regression provide a predicted value for observations with missing values?

Here are the details:

• I have a file with about 10K rows, about 3K have all complete values for all variables.
• I ran a logistic regression (glm) to estimate coefficients and significance for each predictor
• Now when I try to predict the probabilities of the dependent variable using the fitted model, I only get values for the 3K rows that all have values for the independent variables

I would like to have some predicted probability for all 10K rows, is that even possible? How do I do that in R? I have tried na.action=na.exclude/na.pass and that only provides values for the 3K rows.

• If some rows are missing data a common approach is to use data imputation techniques – PolBM Feb 8 '17 at 22:21
• Short answer: NO. – SmallChess Feb 10 '17 at 4:11