[Feature Request]: Additional statistics for the regression module

[gtobin26]

Description

The regression module is quite good but lacks some important features. The first, and most important, is the absence of an Omnibus ANOVA. Second, although the program provides statistical data on categorical variable, the comparisons are fixed, with the first group providing the reference level. Finally, of lesser importance is the absence of the often-used measures AIC and BIC.

Purpose

To add important statistical analyses to allow the regression model to be a standalone program.

Use-case

N/A

Is your feature request related to a problem?

No, except in the sense that alternative sources are needed for the full regression analysis.

Describe the solution you would like

Add options for Omnibus ANOVA, selection of the reference group/level for categorical analysis, and AIC and BIC

Describe alternatives that you have considered

Other software packages

Additional context

No response

[juliuspfadt]

Hi @gtobin26. thank you for the request, and I apologize for the late reply. Let me ask one of our module maintainers for additional help. @Kucharssim, do you know if this is feasible and who could take this up?

[Kucharssim]

Dear @gtobin26,

1. Could you clarify what you are missing regarding the omnibus test? JASP already reports the ANOVA F test for the model (it's the second table in the regression output titled "ANOVA"). Or did I misinterpret your request?

2. The first level of a categorical predictor is indeed used as a reference category in the model. In order to change the reference category, you can reorder the category levels in the data viewer. See this gif for an example of how to enter the menu for reordering the category levels.

3. Adding AIC and BIC sounds quite reasonable to me and I agree it would be a good addition! We can add them in the model summary table.

[gtobin26]

Thx for your response. 1. The Omnibus ANOVA is similar to the ANOVA in the second table (SS etc) but includes the intercept, covariates, and factors. It seems to be a generic term. The Coefficients table includes, I assume, the p-values for the intercept and covariates but not an overall factor (p-values are given for categories within a factor) and that is important. I can send an example if you DM me. 2. I have a workaround by adding zero in front of the categorical predictor so this is a low-level request. 3 Thanks

[gtobin26]

Hi Julius, the layout is given in ANOVA in the R lm package. It includes an option for each of the thee types, I, II, and III.

[tomtomme]

@gtobin26

• If I am not mistaken EJ requests the same omnibus ANOVA here: Omnibus Test for ANOVA #86 - Does the paper there describe, what you want?
• sounds solved by reordering in data view
• BIC and AIC still need to be implemented for Regression

[tomtomme]

@gtobin26
BIC AIC now available via nightlies: jasp-stats/jaspRegression#296