1
$\begingroup$

I thought this might be a fairly simple question, but I've failed gather adequate information on the subject.

So what are the standard pratices to display the results of thousands of regressions? I have been doing some comparisons by looking at graphical averages (an average coefficient value associated with some other characteristic of the dataset, such as geography, e.g. the average value for the west coast is -0.05 if one is running regressions for each city of the united states), but still think I should have some sort summary statistics for the results or another type of result.

$\endgroup$
5
  • 1
    $\begingroup$ Maybe something like a forest plot? Or a meta-analysis for a summary statistic? $\endgroup$ Jan 6, 2021 at 18:14
  • 1
    $\begingroup$ For useful ideas you need to give us some context and details. Many responses and the same predictors? or something else ... you mention cities. Is this paallell regressions for thousands of cities? If so, you could have iteresting data on the cities, and plot regression coefs against them ... see statmodeling.stat.columbia.edu/2005/03/07/the_secret_weap $\endgroup$ Jan 6, 2021 at 18:20
  • $\begingroup$ @kjetilbhalvorsen yes, the same predictors. Thx! $\endgroup$
    – John Doe
    Jan 6, 2021 at 19:38
  • $\begingroup$ @JeremyMiles could you share an example of a forest plot and a meta-analysis of regression coeficients? $\endgroup$
    – John Doe
    Jan 6, 2021 at 19:41
  • $\begingroup$ Sure. Here's one: journals.plos.org/plosone/article/figure?id=10.1371/… . I don't know if you could show thousands, but you'll have a hard time showing thousands of anything on a chart. $\endgroup$ Jan 7, 2021 at 1:31

0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.