search

mpt-network / MPTmultiverse

An R package for the comparison of MPT analysis approaches
3 stars 6 forks source link

Calculation of CIs for no-pooling models #3

Closed mariusbarth closed 6 years ago

mariusbarth commented 6 years ago

Confidence intervals for the no-pooling-models should be calculated in a more comparable way.

mariusbarth commented 6 years ago

Hi @singmann, in mptinr.R, can you have a look at lines 210-246? There, the CIs for the no-pooling approaches are calculated. From my reading of the code, I have the impression that the "CIs" are the quantiles of the distribution of individual parameter estimates, but I'm not 100% sure. As far as I understood, we wanted to calculate CIs similar to se, i.e. just take qnorm(p = .025, mean = est, sd = se). Or am I completely mistaken here?

singmann commented 6 years ago

I have changed this in the code now. The one problem is that CIs can now be larger than 1 or smaller than 0. I think this is just a logical consequence of calculating them based on estimate + x*SE. But I am happy to discuss this. Therefore, I leave the issue open.

mariusbarth commented 6 years ago

I think CIs exceeding the range of [0, 1] are not problematic, as this is what can also happen with the CIs from complete-pooling models.

danheck commented 6 years ago

I agree that this is not a problem for the current project, especially since we are mainly interested in the SE of the mean parameter.

singmann commented 6 years ago

Ok, I will close this issue then.