{[ promptMessage ]}

Bookmark it

{[ promptMessage ]}

Lecture 5_multiple regression_powerpoint2007

# Lecture 5_multiple regression_powerpoint2007 - Lecture 5...

This preview shows pages 1–8. Sign up to view the full content.

Click to edit Master subtitle style 8/5/11 Lecture 5 Forecasting with Multiple

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
8/5/11 Multiple Regression Model Formally, a multiple regression is a statistical procedure in which a dependent variable ( Yt ) is modeled as a function of multiple independent variables ( X1t, X2t, X3t, …, Xkt ). The multiple-regression model may be written as: where β0 is the intercept and the other β i ’s are the slope terms associated with the respective independent variables ( i.e ., the Xit’s ) with i = 1,2,…,k. t kt k t t t t X X X X Y ε β β β β β + + + + + + = 3 3 2 2 1 1 0 (5.1)
8/5/11 Example 5.1 Following the example 4.3 presented in previous lecture, we believe that the US production index is not only influenced by the lagged short-term interest rate, but also follows a stochastic trend. This belief leads us to specify a model as: USPIt = β0 + β 1 USTBRt-1 + β 2 USPIt-1 + ε t (5.2) -A decline in the short-term interest rate will contribute to a decrease in the long-term interest rate, which in turn will stimulate the economy. We anticipate β 1 <0. - USPIt-1 is expected to be positive since the data signify a stochastic trend – with a rise in current production, the trend momentum will carry it to the next period, β 2 >0.

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
8/5/11 Statistical Evaluation We obtain the estimated statistics presented in Table 5.1. Dependent Variable: USPI Method: Least Squares Sample: 1986M01 1999M07 Included observations: 163 Variable Coefficient Std. Error t-Statistic Prob. C 0.137173 0.358204 0.382947 0.7023 USTBR(-1) -0.073664 0.025625 -2.874719 0.0046 USPI(-1) 1.005558 0.003067 327.8179 0.0000 R-squared 0.998671 Mean dependent var 93.90620 Adjusted R-squared 0.998654 S.D. dependent var 12.25531 F-statistic 60117.97 Durbin-Watson stat 2.008455 Prob(F-statistic) 0.000000
8/5/11 The Signs of Regression Coefficients The estimated equation is as follows: USPI t = 0.137 - 0.0737 USTBR t -1 + 1.0056 USPI t -1 , h =1,2,…,12 (5.3) -Are the estimated coefficients are consistent with our prior expectations? -What are the implications? -How reliable are our estimates?

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
8/5/11 Test for Individual Significance Using t-statistic The t -statistic is designed to test the significance of an individual coefficient, such as H0: β 1 = 0 or β 2 = 0. To recall: - Since the t -distribution is symmetrical, we are concerned only with the absolute value of tcalc - If the calculated tcalc -statistic is greater than the critical level (using tcalc > 2 as an approximation), based on ( n -( k +1)) degrees of freedom ( df ), the null hypothesis of no correlation is rejected. - In our example, n = 163 and k = 2, so df =160. The tcalc - statistics are -2.87 for β 1 and 327.82 for β 2. Here both tcalc - statistics are as great deal above the critical value at the 5% level of significance, leading us to reject the null hypothesis of β 1 = 0 or β 2 = 0. / ) ˆ ( 1 1 β β - = calc t (standard error of ) 1 ˆ β (5.4)
8/5/11 Test for Joint Significance Using the F-statistic Test the usefulness of the entire collection of predictor variables by using the F -test. The null hypothesis tested is that none of the Xt variables contributes to predicting Yt by setting the H0 : β 1 = β 2 …= β k = 0. If the null is rejected, then at least one variable contributes to the prediction of Yt . Thus, the alternative hypothesis is that at least one is nonzero.

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
This is the end of the preview. Sign up to access the rest of the document.

{[ snackBarMessage ]}