Include important residual tests for regression analysis: homoscedasticity/heteroscedasticity (Breusch Pagan and White) and normally distributed residuals (SW and KS)

[HBASteens]

Description

Regression statistics do not include tests for homoscedasticity and normal distribution of redisuals

Purpose

Complete regression statistics for conducting standard robustness tests.

Use-case

Saves time and avoids using other applications for simple standard tests

Is your feature request related to a problem?

Above mentioned (simple standard) tests are not covered by statistics in JASP-module for regression.

Describe the solution you would like

Include four statistics: Breusch-Pagan's p-value and White's p-value for testing H0: homoscedastic residuals and Shapiro-Wilk's p-value and Kolmogorov's p-value for H0: normally distributed residuals

Describe alternatives that you have considered

Perhaps other statistics for the same purpose.

Additional context

No response

[tomtomme]

For what its worth, Shapiro is available pairwise and multivariate within the correlation module. But yes, the others are missing. Note, that those tests might not be particularily robust themeselves. Shapiro and KS are under- or overpowered depending on the sample size. Until those tests are implemented - if you like - you could perform those tests graphically with the Plots-Section of the Linear Regression module. If you are not familiar with those plots I am happy to assist. I hope that helps.

[EJWagenmakers]

These would be good to have. Any takers, team?

[HBASteens]

Thank you Thomas, I highly appreciate your prompt response. I feel that adding the BP and/or White statistics for testing homoscedasticity and the SW statistic for testing normally distributed errors to the current regression statistics would enrich the regression functionality of JASP. I haven’t explicitly mentioned in my request that for the homoscedasticty test the distinction between panel data and non-panel data (longitudinal or cross-sectional data) is important. That also goes for the Durbin Watson statistic already covered by the current version of JASP. I hope this helps.

Thank you for all the good work.

Best regards,

Bert Steens

Sent from my iPhone

On 16 Feb 2022, at 16:44, Thomas Langkamp @.***> wrote:



For what its worth, Shapiro is available pairwise and multivariate within the correlation module. But yes, the others are missing. Note, that those tests might not be particularily robust themeselves. Shapiro and KS are under- or overpowered depending on the sample size. Until those tests are implemented - if you like - you could perform those tests graphically with the Plots-Section of the Linear Regression module. If you are not familiar with those plots I am happy to assist. I hope that helps.

— Reply to this email directly, view it on GitHubhttps://github.com/jasp-stats/jasp-issues/issues/1619#issuecomment-1041767807, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AXZQPD2T4XWENK4MMQQHV23U3PA4DANCNFSM5ORVRGHQ. Triage notifications on the go with GitHub Mobile for iOShttps://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Androidhttps://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub. You are receiving this because you authored the thread.Message ID: @.***>

[juliuspfadt]

is this feasible @Kucharssim

[EJWagenmakers]

I'd like to see whether we can implement this. Anybody with time left, people? Maybe this would be a good exercise for the new UD's from Utrecht, to get the hang of it...

[tomtomme]

I am collecting all those test here: https://github.com/jasp-stats/jasp-issues/issues/490 Lets discuss there. Hence closing as duplicate.