The results tell us the null hypothesis of homoscedasticity cannot be rejected for this data. The reg procedure is used for the linear regression and the estat hottest is used to test for the presence of heteroscedasticity. However, both the coefficients are within the confidence intervals of the OLS estimates, and hence, are not significantly different from the OLS coefficients.
Similarly, the coefficient of ed is much smaller than the OLS estimate, for almost all of the quantiles, and is closer to the OLS coefficient around the upper quantiles. The top right-hand plot shows that the coefficient of income is positive over the entire range, and has a larger effect in the median quantile. In the graphs above, the horizontal lines are the OLS estimates which are constants, and hence, do not vary with the location on the x-axis the confidence intervals also appear as dashed lines in each graph. The option scale(1.1) increases the size of the titles in the graphs along the axes. In addition, ci ols olsci produces the 95% confidence interval, the OLS coefficients, and their confidence intervals. The intercept term can also be included in the graph with the option cons. The grqreg procedure produces the graphs from STATA. W: A binary variable representing race (W=1 for whites, and W=0 for non whites) Income: Monthly income of the household, measured in dollarsĮd: Years of education of the head of the household Energy: Total energy consumed, measured in caloriesįat: Total saturated fat intake, measured in gramsĬhol: Total cholesterol intake, measured in milligramsįiber: Total fiber intake, measured in grams
0 kommentar(er)
