The assumptions of ldeveloping the model for estimation is:
- Linear relationship- The relationship should be linear between
independent variable and dependent variable.Linear regression is
sensitive to outliers so it is necessary to check for any
outliers.Thus to test linear relationship use scatter plots.
- Multivariate normality-All the variable is required to be
multivariate and normal.Use histogram or Q-Qplot and check
normality with goodness of fit test(Kolmogorov-Smirnov test).If
data is not normally distributed do a
transformation(log-transformation).
- No or little multicollinearity-When independent variable are
too highly correlated .
- No auto-correlation- Autocorrelation occurs when the residuals
are not independent from each other.Durbin-Watson test. It tests
the null hypothesis that the residuals are not linearly
auto-correlated. Values between 0 and 4. Values around 2 indicate
no autocorrelation.
- Homoscedasticity-meaning residuals are equal across the
regression line can be checked using scatter plot. Goldfeld-Quandt
Test .Non-linear correlation can used to fix homoscedasticity.