I have a numeric dependent variable (drug reports in schools) with three binary categorical predictors - whether schools receive drug training, whether schools receive discipline training, and whether schools have a high or low number of intervention practices. I have made all of these factor variables, and I am now struggling to test the assumptions of my regression analysis. Why is the scatterplot showing vertical lines? Do I need to change my Rcode? I have attatched everything below and any help would be great!
model |t|)
(Intercept) 1.37159 0.06750 20.320
No no vertical lines are fine, but some of these do look a bit wonky. You can quickly run Bartlet's or Levene's test in R to check for homogeneity, but looking at your QQ-plot you probably should check the normality of your data too.
Marius Ole Johansen thank you so much for your response! I have now run the Levene test and I can see there is a violation of the assumption of homoscedasticity. What would you suggest I do to alter these results so that they fit the assumption? Or, what could be said about the weak regression model? Thanks again for any help.
It really depends on how severely we are deviating, many regression models are pretty robust and you could still report your model like it is and highlight assumption violations as "weaknesses of your model". You could also use a robust approximation in your code instead of relying on maximum likelihood/least squares which R uses as a standard (don't recall the syntax on the top of my head, google Rstudio and robust regression and you'll find it) which should be fine with some of the fundamental violations.
I wouldn't call it a weak regression model, two of your βs are significant, and before anyone here flips out about your low R2-value at the end of your output I'll paste in a really good comment from Paetzold from a few years ago regarding low R-squared in social research;
"You need to understand that R-square is a measure of explanatory power, not fit. You can generate lots of data with low R-square, because we don't expect models (especially in social or behavioral sciences) to include all the relevant predictors to explain an outcome variable. You can cite works by Neter, Wasserman, or many other authors about R-square. You should note that R-square, even when small, can be significantly different from 0, indicating that your regression model has statistically significant explanatory power. However, you should always report the value of R-square as an effect size, because people might question the practical significance of the value. As I said, in some fields, R-square is typically higher, because it is easier to specify complete, well-specified models. But in the social sciences, where it is hard to specify such modes, low R-square values are often expected."
Marius Ole Johansen thanks so much for this insight, its been really helpful.
- Badges
- Science topic
- Similar topics
- Social Psychology
- Attitude