Prev: save .mat file
Next: textread
From: Terry Murphy on 18 Mar 2010 09:06
Folks,

Can anyone share sample code to incorporate fixed weights into the mnrfit command?

I'm fitting a model to Medicare data where each observation is a weighted sample. The help information mentions either a sample size of 1 per observation or, taking the row sums of the response matrix Y (n-by-k matrix), where Y(i,j) is the number of outcomes of the multinomial category j for the predictor combinations.

I'm modeling a binary outcome (death) but want each observation to be weighted by a fixed sampling weight provided by Medicare to represent a national sample.
From: Peter Perkins on 18 Mar 2010 09:51
On 3/18/2010 9:06 AM, Terry Murphy wrote:

> I'm modeling a binary outcome (death) but want each observation to be
> weighted by a fixed sampling weight provided by Medicare to represent a
> national sample.

Terry, MNRFIT does not take weights the way that GLMFIT does. But you describe your response as "binary". Can you use GLMFIT instead?
 | 
Pages: 1
Prev: save .mat file
Next: textread