What are the limitations in dichotomising a continuous covariate in simple linear regression?
It's an open question if you have some points please share it!
Dichotomisation of continuous covariate in simple linear regression is a very debatable concept. Although it makes the presentation and interpretation of statistical data easier, it is considered to be statistically non-intuitive because of it's limitations :
1. Dichotomising may work if it creates groups with similar characteristics. But in other hand it results in loss of information.
2. Non-linear relationships such as U-shape associations etc can not be detected through this.
3. Dichotomisation is appropriate only when a threshold effect value truly exists, i.e. if we can assume binary split of the continuous covariate, which creates two relatively distinct but homogeneous groups with respect to a particular statistical output.
4. It is responsible for loss in regression model performance.
Get Answers For Free
Most questions answered within 1 hours.