Step 1: Checking assumptions
The first step is to build forecasting model by checking assumptions of data. There are four assumptions that should
be check which are normality, linearity, heteroscedasticity and multicollinearity. All of the variables in this paper
must be normal distribution. The normal distribution can be seen via histogram graph, plot P-P, plot Q-Q, kurtosis
and skewness. If the distribution of data is not normal, so we need to use transformation.
Then, MLR should have linear relationship between response variable and controlled variables. in regression the
model, we fit is a linear model (‘linear model’ just means ‘model based on a straight line’) (Andy, 2005).
The third assumption in MLR is any data should be free from heteroscedasticity. Heteroscedasticity will happen
whenever there is interruption in the model that not fulfilled. If the important variables in the model are missing,
hence heteroscedasticity will happen. Any model can be check whether there is heteroscedasticity or not based on
Spearman’s rank correlation test.
The last assumption that should be checked in research is multicollinearity. Multicollinearity means situation that
has high degree of correlation between controlled variables. Any analyses can be known the present of
multicollinearity by checking the value of variation inflation factor (VIF). When the value of VIF is less than 5,
hence multicollinearity is not serious. While if VIF is more than 5, then multicollinearity is substantial.
Multicollinearity will be more serious whenever the value of VIF is more than 10.