Trouble with glmfit using binomial distribution and logit

I'm trying to perform logistic regression on a dataset using glmfit. Here is my code:

 N = ones(size(feature));
 initcoeff = glmfit(feature(:, 1:(w-1)), [feature(:, w) N], ...
   'binomial', 'link', 'logit');

w is the width of feature. feature is an array in which each row is an example, and each column is a feature of that example. The last column is a binary vector with the labels for each example as 0 or 1.

When I use 'binomial' as the distribution, I get the following error:

 ??? Error using ==> glmfit at 171
 Y must be a two column matrix or a vector for the binomial distribution.

 Error in ==> comptrain at 28
initcoeff = glmfit(feature(:, 1:(w-1)), [feature(:, w) N], 'binomial', 'link',
'logit');

I'm not sure what to do, because Y (feature(:, w)) is a vector. Using other distributions instead of binomial yields an array index mismatch error from somewhere within glmfit.

Any help would be appreciated. Thanks!

Answer 1

[Tom Lane]

The error message is because you have

N = ones(size(feature));

so it is of the size of the whole feature array. You just want

N = ones(size(feature,1),1);

But even easier is to omit N altogether. The default is N=1.

[Nicole]

Thank you!