slides_Ch4_W[1]

Melissa tartari yale econometrics 38 41 testing

Info iconThis preview shows page 1. Sign up to view the full content.

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

Unformatted text preview: able xj has no partial e¤ect on the dVar y . The hypothesis Ho : βj = 0 all j 1 is an “extreme” exclusion restriction since we test the overall signi…cance of the regression: if we fail to reject Ho then there is no evidence that any of the eVars help explain y (STATA outputs this test by default when we use the command regress) Melissa Tartari (Yale) Econometrics 36 / 41 Testing General Hypothesis: Remarks II For examples of applications of the command test look at the STATA tutorial within the tutorial for the command regress or type help test (also check out the very convenient command testparm). Melissa Tartari (Yale) Econometrics 37 / 41 Testing General Hypothesis: The F Test Here we provide an intuitive overview of the F test. Melissa Tartari (Yale) Econometrics 38 / 41 Testing General Hypothesis: The F Test Here we provide an intuitive overview of the F test. Suppose we want to test the joint hypothesis Ho involving q exclusion restrictions (on q slope coe¢ cients); Melissa Tartari (Yale) Econometrics 38 / 41 Testing General Hypothesis: The F Test Here we provide an intuitive overview of the F test. Suppose we want to test the joint hypothesis Ho involving q exclusion restrictions (on q slope coe¢ cients); What would you start looking at? Melissa Tartari (Yale) Econometrics 38 / 41 Testing General Hypothesis: The F Test Here we provide an intuitive overview of the F test. Suppose we want to test the joint hypothesis Ho involving q exclusion restrictions (on q slope coe¢ cients); What would you start looking at? The sum of squared residuals SSR would seem to provide a convenient basis for testing such hypothesis in the sense that how much SSR increases when we exclude the q explanatory variables tells us something about the truth in Ho . We know that SSR increases when we leave out x 0 s from the model so what really matters is whether such increase is large enough relative to the SSR in the model with all of the variables to warrant rejecting Ho . Melissa Tartari (Yale) Econometrics 38 / 41 Testing General Hypothesis: The F Test The F statistics presented above does exactly this: it combines such information in such a way that the we know its distribution under Ho ! Indeed, we can rewrite it as follow: F= (SSRr SSRu ) /q SSRu /n K 1 (4.37) where: Melissa Tartari (Yale) Econometrics 39 / 41 Testing General Hypothesis: The F Test The F statistics presented above does exactly this: it combines such information in such a way that the we know its distribution under Ho ! Indeed, we can rewrite it as follow: F= (SSRr SSRu ) /q SSRu /n K 1 (4.37) where: SSRr = SSR of the restricted model i.e. the model estimated with the restrictions imposed, Melissa Tartari (Yale) Econometrics 39 / 41 Testing General Hypothesis: The F Test The F statistics presented above does exactly this: it combines such information in such a way that the we know its distribution under Ho ! Indeed, we can rewrite it as follow: F= (SSRr SSRu ) /q SSRu /n K 1 (4.37) where: SSRr = SSR of the restricted model i.e. the model estimated with the restrictions imposed, SSRu = SSR of the unrestricted model i.e. the model estimated without the restrictions imposed, Melissa Tartari (Yale) Econometrics 39 / 41 Testing General Hypothesis: The F Test The F statistics presented above does exactly this: it combines such information in such a way that the we know its distribution under Ho ! Indeed, we can rewrite it as follow: F= (SSRr SSRu ) /q SSRu /n K 1 (4.37) where: SSRr = SSR of the restricted model i.e. the model estimated with the restrictions imposed, SSRu = SSR of the unrestricted model i.e. the model estimated without the restrictions imposed, q number of restrictions Melissa Tartari (Yale) Econometrics 39 / 41 Testing General Hypothesis: The F Test The F statistics presented above does exactly this: it combines such information in such a way that the we know its distribution under Ho ! Indeed, we can rewrite it as follow: F= (SSRr SSRu ) /q SSRu /n K 1 (4.37) where: SSRr = SSR of the restricted model i.e. the model estimated with the restrictions imposed, SSRu = SSR of the unrestricted model i.e. the model estimated without the restrictions imposed, q number of restrictions the statistics F (a RV) is distributed as a F (q , n K 1) Melissa Tartari (Yale) Econometrics 39 / 41 Testing General Hypothesis: the F & the t What is the relationship between F and t Statistics? Melissa Tartari (Yale) Econometrics 40 / 41 Testing General Hypothesis: the F & the t What is the relationship between F and t Statistics? We knew that we could use the t statistics to test Ho : βj = 0tβ = ˆ 1 Melissa Tartari (Yale) ˆ β h j i jx ˆ se βj jx Econometrics tn K1 40 / 41 Testing General Hypothesis: the F & the t What is the relationship between F and t Statistics? We knew that we could use the t statistics to test Ho : βj = 0tβ = ˆ 1 ˆ β h j i jx ˆ se βj jx tn K1 Now we have learnt that there is another way to test the same hypothesis. We call this other statistics Fβ (it has distribution ˆ j F (1, n K Melissa Tartari (Yale) 1)). Econometrics 40 / 41 Testing General Hypothesis: the F & the t What is the relationship between F and t Statistics? We knew that we could use the t statistics to test Ho : βj = 0tβ = ˆ 1 ˆ β h j i jx ˆ se βj jx tn K1 Now we have learnt that there is another way to test the same hypothesis. We call this other statistics Fβ (it has distribution ˆ j F (1, n K 1)). Questions: Melissa Tartari (Yale) Econometrics 40 / 41 Testing General Hypothesis: the F & the t What is the relationship between F and t Statistics? We knew that we could use the t statistics to test Ho : βj = 0tβ = ˆ 1 ˆ β h j i jx ˆ se βj jx tn K1 Now we have learnt that there is another way to test the same hypothesis. We call this other statistics Fβ (it has distribution ˆ j F (1, n K 1)). Questions: How are the two related? Melissa Tartari (Yale) Econometrics 40 / 41 Testing General Hypothesis: the F & the t What is the relationship between F and t Statistics? We knew that we could use the t statistics to test Ho : βj = 0tβ = ˆ 1 ˆ β h j i jx ˆ se βj jx tn K1 Now we have learnt that there is another way to test the same hypothesis. We call this other statistics Fβ (it has distribution ˆ j F (1, n K 1)). Questions: How are the two related? Does inference change depending on which one we use? Melissa Tartari (Yale) Econometrics 40 / 41 Testing General Hypothesis: the F & the t They are related as follows and inference does not change, q tβ = Fβ ˆ ˆ j Melissa Tartari (Yale) Econometrics j 41 / 41 Testing General Hypothesis: the F & the t They are related as follows and inference does not change, q tβ = Fβ ˆ ˆ j j The t statistics is more ‡exible than the F for testing a single hypothesis because it can be used to test against a one-sided H1 . Melissa Tartari (Yale) Econometrics 41 / 41...
View Full Document

This note was uploaded on 02/13/2014 for the course ECON 350 taught by Professor Donaldbrown during the Fall '10 term at Yale.

Ask a homework question - tutors are online