indices of model fit

[Jonas Zaman]

Hi Dominik,

I looked in the manual and on previous post on the forum but have not find the solution to my question.

I was wondering where I could find any indices of model fit (AIC) per subject glm and whether it is possible to get an overall index of model fit across subjects after runing a glm for skin conductance as we want to compare three different model with increasing complexity . I only found the function pspm.predval but I'm not sure how to use it for this end.

Many thanks in advance.

Jonas Zaman

[Dominik Bach]

H Jonas

For a start, please have a look here:
https://sourceforge.net/p/pspm/discussion/help/thread/addfd37e/#5d96

Do update me if you want me to expand on this, or have further questions.

Best
Dominik

[Jonas Zaman]

Hi Domink,

Thanks for you reply. I think I found the solution to my question but just wanted to double check.

So using Y and Yhat I get RSS which I can use to calculate AIC per subject. If I now want to compare three different models across all subjects is it correct that I use the average AIC (across subjects) and compare the averaged AIC's between my three models to compare models?

For i=1:60 % nr of subjects

% with k = number of parameters, n = number of trials

k = length(glm.stats);
n = sum(isnan(glm.Y) == 0);

% get residual sum of squares
RSS = nansum(glm.Y - glm.Yhat)^2;

% Get AIC for a subject

AIC(i) = 2k + n(log(2(pi)RSS/(n))+1);

% Get average AIC across subjects
Average_AIC = mean(AIC);

Many thanks in advance for all the effort and time you spend on helping us.

[Dominik Bach]

This is correct although I'm not entirely sure about the AIC equation, but that is an issue outside PsPM. Dominik