hw1

OOS rmse for cars data, with log

At the link:
oos loop in R

there is an R script to estimate the out-of-sample root mean square error for linear regression used to predict y=price from x=(mileage,year).

The idea of the problem is to see how much it helps to log y.

That is, modify the script to see how much the rmse improve if you “log y”.

At link:
oos loop in python

there is a loop that does exactly the same thing in python.

What does it mean to log y?

• at each train/test split, regress log(y) on x=[mileage,year] using the training data
• predict logy for the test data
• exponentiate the predictions for logy to get predictions for y
• compute the rmse

Before you do this, plot mileage vs. log(y=price) and year vs. log(y=price).
Based on the plots does logging y look like a good idea?

Only one way to find out !!!

Use rmarkdown or jupyter notebook to present your code and results.