{[ promptMessage ]}

Bookmark it

{[ promptMessage ]}

ECON301_Handout_12_1213_02

# 2 in our example t k 113 5 17 3 in our example p3

This preview shows pages 12–16. Sign up to view the full content.

2 In our example T-k-1=13-5-1=7. 3 In our example p=3.

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

View Full Document
ECON 301 - Introduction to Econometrics I May 2013 METU - Department of Economics Instructor: Dr. Ozan ERUYGUR e-mail: [email protected] Lecture Notes 13 Hence, the steps of predictive test can also be written as follows: Step 1. Estimate the equation over the complete sample period [i.e., estimate equation (1)] and retrieve the sum of squared residuals, R S . Step 2. Estimate the equation over the larger sub-sample and retrieve the sum of squared residuals, * U S . Step 3. Calculate the following test statistic:     * * 1 / /1 RU U S S p Q S T k  where 1 1 1 Tk is the degrees of freedom 4 in unrestricted model (1*). The calculated Q statistic is distributed as F with [ T 2 , 1 1 1 ] degrees of freedom: 11 ,1 p T k QF .Reject the null hypothesis of parameter constancy if p T k . 2. Unknown Breaks: Testing for Break Points A break is another name of a structural change point in data. Breaks can arise either from a change in the population regression coefficients at a specific date or from a gradual evolution of the 4 In our example T 1 -k-1=10-2-1=7.
ECON 301 - Introduction to Econometrics I May 2013 METU - Department of Economics Instructor: Dr. Ozan ERUYGUR e-mail: [email protected] Lecture Notes 14 coefficients over a longer period of time (Stock and Watson, 2012, pp. 598-610). One way to detect breaks is to test for discrete changes, or breaks, in the regression coefficients. How this is done depends on whether the date of the suspected break (the break date) is known. In some applications you might suspect that there is a break at a known date. We have seen that (last two lectures), if the date of the hypothesized break is known, then the null hypothesis of no break (no structural change) can be tested using ACOV Chow test and Predictive Chow test. Chow suggested that the predictive test for the case where second subsample size ( T 2 ) is less than k +1. In this case the regression equation cannot be estimated with the second sample and thus the analysis of covariance test (ACOV test) cannot be applied. In this case only the predictive test can be used. He also suggested that the predictive test can be used even when second subsample size ( T 2 ) is larger than k +1 [i.e., T 2 > k +1] but that in this case the ACOV Chow test should be preferred because it is more powerful (Maddala, 1992, p.176) On the other hand, often the date of a possible break (structural change point) is unknown or known only within a range. In this case three tests can be used to determine the break date:

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

View Full Document
ECON 301 - Introduction to Econometrics I May 2013 METU - Department of Economics Instructor: Dr. Ozan ERUYGUR e-mail: [email protected] Lecture Notes 15 (1) Quandt likelihood ratio (QLR) statistic (Quandt, 1960) 5 (2) CUSUM and CUSUMSQ tests (Recursive Residuals) In this semester of ECON301 we will only see CUSUM and CUSUMQ tests. A. CUSUM and CUSUMSQ tests We may have no strong prior knowledge of specific structural breaks. Another method used to avoid arbitrary breakpoints (structural change points) is based on recursive estimation . This involves estimating the equation by OLS; starting with a small sample and allow the data period to increase by one period each estimate. It therefore produces a time series of estimates for the coefficients. These can be plotted with confidence intervals.
This is the end of the preview. Sign up to access the rest of the document.

{[ snackBarMessage ]}

### Page12 / 19

2 In our example T k 113 5 17 3 In our example p3 ECON 301...

This preview shows document pages 12 - 16. Sign up to view the full document.

View Full Document
Ask a homework question - tutors are online