Extract details on Pareto K after fitting a model

[franzis...@gmail.com]

Hi,

thanks for the great package and the help forum. 

When fitting models I often experience high Pareto ks. How can I extract information for Pareto k, other than the warning that is displayed?

 I would like to get something like this as available from the standard Stan code:

Computed from 12000 by 20 log-likelihood matrix Estimate SE elpd_loo -1875.5 83.0 p_loo 32.6 2.0 looic 3751.0 166.0 ------ Monte Carlo SE of elpd_loo is NA. Pareto k diagnostic values: Count Pct. Min. n_eff (-Inf, 0.5] (good) 0 0.0% <NA> (0.5, 0.7] (ok) 0 0.0% <NA> (0.7, 1] (bad) 16 80.0% 21 (1, Inf) (very bad) 4 20.0% 9 See help('pareto-k-diagnostic') for details.

I might be obvious, but I seem to be unable to extract these bits of information.

thanks a lot!

Franziska 

[Eunhwi Lee]

Hi Franziska,

You can utilize the 'extract_ic' function to get the Pareto k diagnostic values you inquired about.

For a comprehensive understanding of this function, please visit the following link: https://github.com/CCS-Lab/hBayesDM/blob/develop/R/R/extract_ic.R.

Please do not hesitate to reach out if you need further assistance.

Best,

Eunhwi

2024년 2월 13일 화요일 오후 2시 38분 55초 UTC+9에 franzis...@gmail.com님이 작성: