Question about repeated measures Anova and multiple regression

[miau]

This is for my coursework. I have two problems. The first one.

Homework Statement

I have two time messures. Before and after a scheme was introduced.I have to answer a question if it made a statistically significant difference or not. And also as it was introduced in 3 places I need to know if it worked equally well in all of them.

Two time meassures are on the interval scale, place is a categorical variable place 1, place 2 and place 3.
The major problem is that data is not normally distributed (and it is impossible to fix that, I tried various data transformations), so my question is which non-parametric equivalent test should I use?

Homework Equations

To be honest I can't decide if it is a one way repeated measures anova or mixed factors repeated measures anova as I am still new at this topic. I just keep changing my mind and performing these tests again and again (parametric versions, although I am aware I should go with non-parametric test)

The Attempt at a Solution

As I mentioned before I tried tranforming the data in order to make the distribution normal but I have never (so far ) seen anything so far from the normal distribution pattern and nothing is working.
Another question:

Homework Statement

I have two IV and 1 DV (all are interval) and I need to build a multiple regression model to be able to predict DV from these two IVs . However, residuals are not normally distributed also there is heteroscedascity.

How can I fix this, should I use general linear model (via SPSS) to build my equation instead? How? So far I only dealt with data which met the assumptions.

Homework Equations

I need to use multiple regression, also to find which of the two is better predictor

The Attempt at a Solution

I tried data transformations but how do I know which variable to transform and which transformation to use? Also I tried using general linear model but I am not sure if that is aprropriate.


I didn't give much details because I don't think I can do that. I hope the information I provided is enough to understand the types of problems I am facing.

Thank you.

 

[LightHero]

The solution to your first problem is to use a non-parametric test, such as the Kruskal-Wallis test, to determine if there is a statistically significant difference between the before and after measurements. This test can also be used to compare the difference in performance across the three places. For the second problem, one solution is to use the Generalized Linear Model (GLM) instead of the multiple regression. GLM allows for the testing of non-linear relationships and can be used to account for heteroscedascity and non-normal residuals. You can also try using polynomial terms or transformations of the independent variables to reduce heteroscedascity and make the residuals more normally distributed.