In: Statistics and Probability
Question
a) State two possible consequences of not including an intercept in an estimated regression model
b) When is heteroscadasticity said to occur in a regression model?
c)State the steps involved in conducting the Goldfeld-Quandt test
d)State the consequences of using OLS in the presence of heteroscedasticity
e)State two solutions for heteroscedasticity
f)State the conditions that must be fulfilled for Durban -Watson test to be a valid test
g)Why mighy lags be required in a regression model?
h) State two solutions to the problem of multicollinearity
a) First is that the regression line gets shifted to the origin and thus all the values got decresed by the value of intercept. if the intercept is positive there is constant decrease in value of y for all values of x. if intercept is negative there is constant increase in the value of y for all values of x.
Second is , The value of y when x is 0 gets lost, thus there is no information about what values y can take without x.
b) When the variance of the population is not constant than the heteroscadasticity is said to occur in regression model. when there is large range of values between highest and the lowest value then also it occurs.
when the error variances vary with a proportionality factor than also it occurs.
c) Steps are :-
1.Order the data in ascending order, generally ordering by magnitude is done.
2. It is recommended to divide the data into three parts., top , middle and the bottom part.
3 observations are dropped in the middle part.
4. Run separate regression analysis on the top and bottom parts that is on the other two parts apart from the middle part. for both the regression equations the residual sum squares is calculated.
5. Calculate the ratio of the Residual sum of squares, where RSS2 is the set of high values(upper part) and RSS1 is the set of low values (bottom part) : (RSS2/df)/(RSS1/df).
6. now apply the conventional decision rule of the f-test. the larger the value opf F the more are the chances that variances ate different.
d) OLS assumes that all the residuals are drawn from the population with the constant variance, where as there is difference ibn varisnce in case of heteroscadasticity. It makes the coefficient estimates less precise.