Drawbacks of Regression Analysis
- Multi-collinearity- these are cases where two or more explanatory variables in a regression model are highly correlated, which makes it difficult to differentiate the individual effects on the dependent variable. This drawback usually leads to the estimated OLS coefficients being statistically insignificant and even have wrong sign.
- Heteroscedasticity- This drawbacks sets in when the statistical assumption that the variance of the error term being constant for all independent variables is violated and found not to hold. It leads to inefficient estimates of the standard errors, including incorrect statistical tests and confidence intervals.
- Regression models are sensitive to outliers and negligible data- with outliers having univariate or multivariate basis, regression models such as liner factors in such data causing huge statistical errors.
- Most regression models use incomplete data, and falsely conclude that a correlation is a causation. This drawback can lead to inferences being made based on related but not proved cases.
- Linear regression only focuses on the mean of the dependent variable and ignore the rest of the variables, hence the information and conclusion in not complete. For instance, in the birth of an infant and the maternal; features such as age, linear regression looks only at weights of babies born to mothers different ages, without considering other extreme independent variables.