2
$\begingroup$

I am trying to estimate a multinomial logit model using an offset variable with the mlogit package of R

Using the syntax here [https://stat.ethz.ch/R-manual/R-devel/library/stats/html/formula.html] it seems that the correct way to do this is to use something like this

v <- mFormula(choice ~ gc + log_gc + offset(log_attr) + 0)

which you would then pass to the model estimation as

model <- mlogit(v, mlogit_data)

Printing the coefficients seems to indicate that the parameter has not been estimated

model$coefficients
          gc       log_gc 
-0.003023088 -0.477115780 
attr(,"fixed")
    gc log_gc 

However these parameters are exactly the same as the parameters I get if I don't include the variable log_attr at all... ie

v_no_log_attr <- mFormula(choice ~ gc + log_gc + 0)
model_no_log_attr <- mlogit(v_no_log_attr, mlogit_data)
model_no_log_attr$coefficients
          gc       log_gc 
-0.003023088 -0.477115780 
attr(,"fixed")
    gc log_gc 
 FALSE  FALSE 

Given that the data has significant variation in this variable, I feel that this is unexpected and that the mlogit package is simply ignoring the offset variable.

variation_in_data

Is this possible to do with this package? Do I have a misunderstanding of how to perform the estimation?

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.