A regression model built by a dataset could sometimes demonstrate a low quality of fit and poor predictions of individual observations. However, using the frequencies of possible combinations of the predictors and the outcome, the same models with the same parameters may yield a high quality of fit and precise predictions for the frequencies of the outcome occurrence. Linear and logistical regressions are used to make an explicit exposition of the results of regression modeling and prediction.
An earlier version of this article mistakenly replaced ellipses with a capital letter K. This has been corrected.
Lipovetsky, S. (2019). Regression modeling and prediction by individual observations versus frequency. Journal of Modern Applied Statistical Methods, 18(1), eP2692. doi: 10.22237/jmasm/1556669100