Final answer:
Transformations of responses or predictors in linear regression are usually done before running the analysis to satisfy the model's assumptions. Sometimes after initial analysis, further transformations may be considered as part of model refinement.
Step-by-step explanation:
The transformation of responses or predictors in a linear regression should generally be done before running the linear regression analysis. Transformation before analysis can help meet assumptions of linear regression such as linearity, constant variance, and normality of errors. If the relationship between the response and predictors is not linear or if the variance of the errors is not constant, a transformation can help rectify these issues. By transforming beforehand, the model fitting can be more effective and resulting interpretations more meaningful.
However, it can be necessary to transform after the initial analysis if the residuals suggest that transformations could improve the model. Yet, this is usually part of an iterative process for refining the model rather than a primary strategy.