Question

DR 252: Applied Statistics Total ring 2019 Quiz 13 Name: Karun Dor yon y 8, 2019 -- we have n observations (zn, yī), (z21, บ2), . . . , (Zn1.Yn), where the response variable yi takes on the values either 0 or 1. We generate another training set of n observations {(x11,212, yı), (z21, Χ22W2), . . . , (4m Ind, yn)), where in each observation zil and Vi are the same as in the previous set and then t2 is a duplicate of T. Does logistic regression suffer from having this additional duplicate variable? Justify your answer. (10 points) 1 , Proo

Answer 1

No, the logistic regression does not really suffer from having this duplicate set of regressor variables.

This is so because the replicates are being introduced for all of the observations in the dataset. So, all the regressor variables get weighted as before introducing the replicates. That is, there is no difference in the weights of the regressor variables before and after introducing the replicates. However, if the replicates would have been introduced for a subset of the data points, it might have affected the coefficients and introduce bias in the model