8.28 Coursera Regression Models Project Comments

Apparently, I don't think that I can get full credits on this subject because the uncertainty is not fully explained by giving confidence intervals. However, these comments do help me to improve and remind me keep learning regression models.

 

peer 1 → I believe the student did a good job in selecting the key predictors. Also the student provided a good explanation of how the variables selected and how the final model was derived. I believe the student left out some key steps but I can he did not want to have more than six pages.

The last sentence makes me laugh. But I need to learn how to create pdf within page limits. My way of compiling pdf is not the best.

 

peer 2 → Strengths: Nice structure and analysis. I came up with the same conclusions via slightly different methods. Weaknesses: It doesn't appear that you used knitr to generate this PDF. The code appears to be more of a word document format rather than output of a ".Rmd" processed by knitr into a PDF as required. Additionally, you omitted a bunch of steps on how you arrived at your optimal model.

I should include key steps, and probably I need to take "Reproducible Research" to learn how to make pdf using knitr.

 

peer 3 → I think the final model suffered a little by focusing on r^2 and the p-values. I think that testing the assumptions behind linear regression should also be considered because the final model has a few violations with unequal variance of the error term and a non-normal distribution. A really good thing I saw is that the final model doesn't appear to have any issues with multi-collinearity, which was a big issue not addressed in the other 3 projects I saw. 

This is the most specific comments of the three. It points out some weaknesses of my strategy, and I believe the person has knowledge in math more than I do.