GLM - Continuous Target

Question 1

Weights

Answer 1

Variable of positive constants that:
-measure the relative importance of each observation
-scale the variances of the target for all observations
-are intuitive as integers, making the target more interpretable as an average

Weights are w_i’s that come from replacing phi with phi*w_i in the log-likelihood function l(b). Weights impact how B is estimated, attributing more importance to the observations with greater weight.

If the target is recorded on a total basis, it should be transformed to be on a per weight basis before modeling

weights should be used if there is reason to believe each observation/row of a dataset should not be weighted/treated with equal importance in estimating the coefficients.

Question 1

Using Gamma/Inverse Gaussian

Answer 1

Adding 1 (or another small positive value) to Time.to.resolution is necessary since the gamma and inverse Gaussian distributions do not support 0 values, which do exist in the Time.to.resolution variable. Using the log link function ensures that the predictions are positive.