I would like to compare AIC values of a single growth curve fit to an entire dataset vs. the combined AIC values of two growth curves (one fit to male data points in blue and one fit to female data points in red).
Is this possible?
1 Answer
$\begingroup$
$\endgroup$
3
Not really. The AIC is a value that depends on the likelihood value, which means it's only comparable for curves fitted on the same data. On this case, you divided your data, generating two likelihood values, and they're not comparable to the likelihood value of the whole dataset.
-
$\begingroup$ Ok. Basically what I am trying to do is take a dataset and determine if one curve or two curves better explain the data, while punishing the option with more curves. This is what brought AIC to mind, but it seems like it might not be applicable. $\endgroup$ Commented Nov 22, 2017 at 17:25
-
$\begingroup$ Would there be another way to approach this? $\endgroup$ Commented Nov 22, 2017 at 17:25
-
1$\begingroup$ One useful and common approach is to check the sum of residuals squared. The "best" model is the one that produces the smallest values for this quantity. $\endgroup$– Bruna wCommented Nov 22, 2017 at 18:33