Further updates to @ruchikasonagote's PR #141 have been integrated into release-1.6.6. Functionality is still missing and more work is required.

/assign

Hello there,

Some points are not very polished. I'm working to solve the problem of deviance that occurs when the distribution is not binomial (see the following test prompt)

***** test
rand ("seed", 1);
X1 = rand (50, 1);
X2 = rand (50, 1) * 0.5;
b_true = [0.4; 1.5; -0.7];
mu_true = exp (b_true(1) + b_true(2) * X1 + b_true(3) * X2);
randp ("seed", 1);
y = poissrnd(mu_true);
[b, dev] = glmfit ([X1, X2], y, "poisson", "link", "log");
assert (b(1), b_true(1), 1);
assert (b(2), b_true(2), 1);
assert (b(3), b_true(3), 1);
assert (dev < 60, true);
!!!!! test failed
corrcoef: X and Y must be the same size

Thinking in the future, I want to give glmfit the full compatibility with other distributions like p-dist.

Thank you for your dedication

Your testing code above works properly with the latest dev sources bafore merging your PR. See my comments on your PR.

pkg load statistics
>> X1 = rand (50, 1);
>> X2 = rand (50, 1) * 0.5;
>> b_true = [0.4; 1.5; -0.7];
>> mu_true = exp (b_true(1) + b_true(2) * X1 + b_true(3) * X2);
>> y = poissrnd(mu_true);
>> [b, dev] = glmfit ([X1, X2], y, "poisson", "link", "log");
>> assert (b(1), b_true(1), 1);
>> assert (b(2), b_true(2), 1);
>> assert (b(3), b_true(3), 1);
>> assert (dev < 60, true);
>>