Hello,
I have a non-linear regression model of a first order ODE in which 6 dependent variables are estimated parallel. If I want to calculate an information criterion such as Akaike, can I simply sum up the individual variance of each dependent variable's residual or shall I calculate the variance of all dependent variable's residuals at once?
Best,
Dennis
You have six different models. They may be compared on the basis of the AIC in which case the criterion may be calculated for each model. Then the model with the least AIC is the best of them. It makes no sense to want to calculate the variance of the dependent variables all at once.
I wouldnt speak from six different models, since I estimate the parameters of all dependent variables at once and they partly use the same parameters. And my goald would be to compare results of such combined estimations of subsets from the observations and alternative number of included parameters.
- Badges
- Science topic
- Similar topics
- Psychology
- Mental Processes
- Learning