I am working on a database of 6500 patients with 150 variables. There are variables that are complete, however, certain variables are missing by 5%-15%. So I am conducting multiple imputations on this dataset. Initially, I ran the default command with mice with which I got an error: Error in solve.default(xtx + diag(pen)) :
system is computationally singular: reciprocal condition number = 2.00305e-16
I changed the codes to mice(data1, m=1, maxit=500, method='cart', seed=500). I am wondering:
- Is it appropriate to use m=1 or it should be at least 5?
- When I do the linear regression, I have treatment variable "cg" and outcome variable as "fs". Which variable should I be using for modeling:
fit <- with(imp, lm(??? cg or fs ~ LOC + AGE + gender + TEMS+...)
Thank you in advance.