Use matrix scatter plot and visually decide if a linear model is appropriate 

2. If the relation is not linear, transform the data accordingly.

• Try logarithm, exponential, square root, square, etc., for Y and/or X until you see a linear relation. You only need to report what is the transformation chosen,

3.Use an OLS model for the original and transformed data if required.

• Evaluate if the OLS assumptions are met: normality of errors centered around zero, equal variance, etc..., for the original data and transformed data if appropriate.
• Comment how the transformation impacted the different assumptions. (This should be done only by looking at the output diagnostic charts created by the software)
• If datasets have outliers, remove the outliers and see the effect in the model (slope, intercept and R-square)

https://drive.google.com/file/d/1ut12k9LlrKA9I9gL8Bxau4y83xU7jH9d/view?usp=sharing