*3.4. Data Diagnosis*

To ensure linearity and to avoid the outlier problem, we ran primary regression on the independent variables to obtain residuals and estimated values for dependent variables. We then plotted the residual and estimated values on the residual-versus-fitted graph to detect linearity. Then, we corrected the outlier problem by winsorizing data at 1% and 99% tails. After winsorizing, we had 7055 observations. Figure 1 confirms linearity after data winsorizing.

**Figure 1.** Residual-versus-fitted plot after data winsorizing. Source: authors' construction.
