centrality indeces

Larisa Morosan

unread,

May 16, 2024, 2:21:51 PM5/16/24

to gimme-r

Hello all,

I run a GIMME network on EMA data and I want to compute the overall strenght/infleunce of each node for each particiapnt, so that it can to be used in a follow-up analysis.

Is there an option to obtain the strenght of each node automaticlally after running GIMME? Is there an option to calculate the expected influence or a relative influence index (e.g. LMG) of each node?

Furthermore, in summaryFIt document, some of the R2 are bigger then 1 (e.g. 54), which seems strange. Do you have any ideea why this values are so big (note: the status of the model is "converged normally")?

Thank you in advance for your help!

Larisa Morosan

Katie Gates

unread,

Jun 25, 2024, 12:01:13 PM6/25/24

to gimme-r

Hi Larisa,

Thanks for letting me know of the R2 issue - it was a part of the "high estimates" issue and has been resolved on the last update (as well as the high beta estimates, which I know your data had!).

For your other question, others would be more knowledgeable about how to come up with network and influence measures from VAR output such as what you obtain from gimme.

Best,

Katie

Mercedes Woolley

unread,

Jul 3, 2024, 12:10:41 PM7/3/24

to Larisa Morosan, gimme-r

Hi Larisa,

To help with your centrality question, there are a couple ways to go about it and it may depend on your research question. For some work we just did, we were interested in interpreting individual level networks, so for each individual we calculated out-strength centrality by summing the edge weights (from the indivPathEstimates file), specifically for outdegrees, in order to interpret the "influence" each node had on the network. Happy to share more details about what we did, just reach out.

Mercedes

--
You received this message because you are subscribed to the Google Groups "gimme-r" group.
To unsubscribe from this group and stop receiving emails from it, send an email to gimme-r+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/gimme-r/c25146bb-f80d-4144-aed4-ba863340f692n%40googlegroups.com.

Bringham et al. 2019.pdf

Larisa Morosan

unread,

Jul 3, 2024, 12:10:50 PM7/3/24

to gimme-r

Hello Katie,

Thank you a lot for your answer!

Unfortunatly, the estimation still results in high beta values. Do you have an ideea why and how this is still possible after the update of the package?

Thank you for your hep!

Best,

Larisa

Katie Gates

unread,

Jul 3, 2024, 12:26:12 PM7/3/24

to gimme-r

Hi Larisa,

Could you please check the status of those individuals with high betas? This will be in the $fit dataframe (if output is saved as an R object) or summaryFit.csv. If it is "unstable", then these individuals likely should not be used.

If the status indicates "converged normally" for these individuals, I'm curious how high is "high"? We have seen that sometimes values over 1 still result in a stable and ok solution.

Best,

Katie

Larisa Morosan

unread,

Dec 10, 2024, 10:30:05 AM12/10/24

to gimme-r

Hello Katie,

I guess it is obvious by now that I totally missed your message :-/

Thank you very much for your answer!

The status of all participants was "converged normally". The highest values was -4.16 and 1.66. As they were not so extreme, we decided to include them in the analyses.