Hello,

I am trying to analyze some data from a choice experiment. I want to use the "mlogit" package. However, I have an error in the MNL function. I tried to find solutions in specialized sites and in forums but without success.

My choice experiment is the election of three options with three attributes (with three levels 0 for the status quo, 1 and 2), plus a payment variable to improve the other three. In this way, each individual has eight sets with three options with different combinations of attributes (option 3 the status quo 0 for all attributes). However, when I run the modedl with the 3 attributes (as factors) and the payment, this error appears:

Error in solve.default (H, g [! Fixed]):

system is computationally singular: reciprocal condition number = 1.44666e-21

I read that it could be a problem of multicollinearity, because the error appears only when I put the attribute "REC", but it also happens when I do the function only with this variable, and does not appear with the other two attributes or with the payment (alone or together). So I think this is a problem with this variable, but I don't know why it would be a problem if it is equal to the other two attributes.

if someone comes up with an idea I would appreciate it