The points should create a corridor that looks like a logarithmic, square root, inverse etc graph...
For simple linear regression:
Given a scatter plot, in general, what are some features of a plot that suggest a transformation(log, sqrt, inverse) of both variables(dependent and independent) are necessary? (should reference 1 concern for each variable)
I know that we use transformation to overcome problems with non constant variance, non normality, and non linearity.
But just from a scatter plot, what should i look for?