Mckinsey Interview Questions | Mean Imputation

Question

Is the mean imputation of missing data acceptable practice? Why or why not?

in progress 0
Dhruv2301 4 years 2 Answers 1226 views Great Grand Master 0

Answers ( 2 )

  1. It is acceptable when the missing value proportion is not large enough.
    But, when the missing values are large enough and you impute them with the
    mean, the standard errors will be lesser than what they actually would have been.
    Small standard errors can lead to small p-values and this can create problems for us,
    because some variables will start appearing significant, which are ideally not significant.

  2. If the numerical column follows Gaussian distribution imputing with mean value makes sense, otherwise we prefer median . If the distribution is skewed, your mean will vary and be influenced by outliers. To avoid that, we use median imputation

Leave an answer

Browse
Browse