Re: Binary Logistic Regression

The glmfit function wants to operate on columns. Make sure x1,x2,x3,y are
all column vectors, and then xfull will be a three-column matrix.

The second argument to chi2cdf in this example is a 3 because there are
three columns of the X matrix.

-- Tom

"jacksparrow " <shobhitjain2166@gmail.com> wrote in message
news:kjuc4u$n1b$1@newscl01ah.mathworks.com...
> Hey Phoebe!
> I am getting the same problem. Did you get the correct solution now?
> Please share.
>
>
> "Phoebe " <phoebe_62@hotmail.co.uk> wrote in message
> <k1ij0v$5gt$1@newscl01ah.mathworks.com>...
>> Trying to perform binary logistic regression with one y and 3 x's. Code
>> is as follows so far;
>> clear
>> x1=[....];
>> x2=[....];
>> x3=[....];
>> xfull=[x1 x2 x3];
>> y=[0 1 0 ...etc];
>> [bfull,devfull]=glmfit(xfull,y,'binomial','link','logit');
>> n1=sum(y);
>> n=size(y,1);
>> n0=n-n1;
>> G=-2*(n1*log(n1)+n0*log(n0)-n*log(n) - dev);
>> pval=1 - chi2cdf(G,3)
>> bfull
>> devfull
>> n1
>> n
>> n0
>> G
>> pval
>>
>> I have a few queries;
>> 1. Matlab says my y must be a 2 column matrix or a vector which it is
>> inputted as so i dont know why it doesnt recognise this?
>> 2. Within the pval equation it refers to (G,3) although in some cases
>> this number is a 1 not a 3?