STATA mle function for Mcfadden MNL model

[Jordan Foster]

I am estimating the covariates for an aggregate market data set. I am aware of the pre-existing STATA packages out there like mlogit, clogit etc.., however they do not fit my case as I am working with aggregate and not individual level data. I have tried writing out a MLE function but it does not seem to be working

program define mcfadden
          args lnf Xb
          quietly replace `lnf' = `Xb' - ln(1+sum(exp(`Xb'))) 
     end

My dependant variable is market shares and my independant are product characteristics like price and product attributes. So I am estimating the likelihood as given below



I need help on knowing what's wrong with my code.

[Nick Cox]

Cross-posted at https://stackoverflow.com/questions/...dden-mnl-model

sum() is for cumulative or running sums across observations, which is not what you seek here. Other people will be more familiar with this model and may be able to advise further. You may need to define your terms however.

[Maarten Buis]

See ssc desc fmlogit