13.9 Multiple Linear Regression: Transformations

Поділитися
Вставка
  • Опубліковано 17 гру 2024

КОМЕНТАРІ • 6

  • @cengiz2645
    @cengiz2645 2 роки тому

    Youre so underrated! Great video, thank you so much- ive been stuck on an important exam question. I hope your channel continues to grow so you get the recognition you deserve!

  • @nashkmr
    @nashkmr 3 роки тому

    Thanks so much for the great video. Helped me interpreting the log transformed models.

    • @ShawnJanzen
      @ShawnJanzen  3 роки тому

      Thanks Naresh! Glad you found it useful.

  • @couragee1
    @couragee1 3 роки тому

    thank you so much!

  • @hr-xd8bf
    @hr-xd8bf 4 роки тому +2

    Hey Shawn, thanks for the comprehensive vid! One question about the log transformation for multiple regression models: my model includes two independent variables, of which only one (intensity) is not normally distributed. Can I just log this iv intensity and leave the other iv as it is or do I have to transform both of them?

    • @ShawnJanzen
      @ShawnJanzen  4 роки тому +1

      Hi hr, glad you liked the vid! One nice thing about regression is that you have control over how you use your IVs. You can choose to log or not log (or similarly modify or not) any of your variables individually. So while you can log both IVs, there is not requirement to log both if you only want to log the intensity IV. Also, while there is no assumption that the DV is normally distributed for linear regression (residuals should be though), you can separately choose to log the DV or not. In fact, while interpreting log results are not terrible compared to other types of transformations, I usually find it easiest to explain regression coefficients on untransformed (original) IVs. Depending on how you are using software for your regression, you can set it up to log each specific variable in the regression; or choose to make a new log variable for each IV (and/or DV) before doing the regression and then use the logged form instead of the original in the regression. I usually recommend looking at models with and without the logged form variables and see what changes in terms of each variables and overall model fit (see my model fit testing vid). Good luck!