logistic regression

I am trying to use logistic regression, but when i execute the following line of code

betaHatNorm=mnrfit(data,class);

my betaHatNorm contains mostly Nan, any resaon it might do so? it also throws a warning regarding matrix

Warning: Matrix is singular to working precision.

Answer 1

[Peter Perkins]

What most people mean by "logistic tregression" is a binomial response: number of "successes" vs. number of "failures". Unless you specifically need to fit a model for a multinomial response (more than just two response categories), I'd suggest you use glmfit.

[Nirmal]

Thank you for the response. What is the reason for using glmfit?

[Peter Perkins]

"Unless you specifically need to fit a model for a multinomial response ..." mnrfit is both more and less general than glmfit, but it is simpler to use glmfit for (binomial) logistic regression.

[Runcong Kuang]

Hi @Peter Perkins,

I am doing multinomial response using mnrfit.

I am getting the same warning.

How can I fix this?

Thanks.