search

SachaEpskamp / psychonetrics

An R package for Network Psychometrics
43 stars 5 forks source link

Confidence Intervals - illegible and most of it not significant #35

Open usr7777 opened 4 months ago

usr7777 commented 4 months ago

Hello Sacha,

I am trying to get the confidence intervals plots: CI_beta.pdf CI_temporal.pdf

The black lines apparently represent a p-value > 0.05. Hence, given these two graphs above, most of it is not significant, right? Besides, the graphs are pretty illegible even when I extra zoom in. Is this normal, please?

SachaEpskamp commented 4 months ago

This looks quite usual; there are quite a lot of significant edges. Did you use thresholding when extracting the matrix to plot? e.g., getmatrix(model, "PDC", threshold=TRUE)?

usr7777 commented 4 months ago

Yes, to extract the temporal network (getmatrix(mod, "PDC", threshold = TRUE)) and the contemporaneous network (getmatrix(mod, "omega_zeta_within", threshold = TRUE)), I used thresholding. Only for the between-person network, I used the saturated model (getmatrix(mod, "omega_zeta_between")). My networks looked like that after thresholding the temporal and contemporaneous networks only (which also seemed a bit strange to me, but I guess maybe that's how it should be): thresholded_network.pdf

All the 3 networks, in the saturated form, looked like that, which is worse: Networks_trial1.pdf

For the confidence intervals, I was able to extract only for the temporal and contemporaneous networks, for the betwee-person, I was getting errors.

Look at CIs:

CIplot(mod, "beta") CIplot(mod, "omega_zeta_within") CIplot(mod, "omega_zeta_between") Error in data.frame(edge = edges, est = est, p = p, lower = ci_lower, : arguments imply differing number of rows: 1, 0

The model fit seemed okay though: Model fit.txt