Lecture 9

Lecture 9 - Introduction to Diagnostic Testing In a...

Info iconThis preview shows pages 1–4. Sign up to view the full content.

View Full Document Right Arrow Icon
1 Introduction to Diagnostic Testing In a regression equation the N observations generate • K regression coefficients • N residuals with (N - K) distinct components (because Σ e i = 0 etc) These residuals should not have a systematic pattern Diagnostic statistics attempt to detect such patterns, which may indicate misspecification Misspecification leads to biased coefficient estimates with undesirable properties However diagnostic tests have low power if the regression equation has high variance The assumptions behind OLS The assumptions needed for OLS to be BLUE are 1E [u i ] = E[y i - Σβ j x ij ] = 0 Y is a linear function of regressors 2E [ (u i ) 2 ] = σ 2 disturbances have common variance σ 2 3E i u j ] = 0 if i j Assumption of independent disturbances – implied by random sampling Tests of this assumption are more relevant when using time series data: covered in Lecture 15 Figure 1 Nonlinearity and Structural Breaks 0 1000 2000 3000 1 6 11 16 Actual Values Ful Sample Regression Sample 1 Regression Sample 2 Regression
Background image of page 1

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

View Full DocumentRight Arrow Icon
2 Figure 2 Nonlinearity and Residuals -300 -200 -100 0 100 200 1 6 11 16 Testing Assumption 1 - Linearity Suppose the true model is not linear (violating Assumption 1) Y = α + β f(X) where f(X) = γ 1 X + γ 2 X 2 The linear regression Y = α + β X is misspecified because it omits X 2 Examples Linear rather than quadratic trend Cobb-Douglas rather than CES production function Why test for linearity? Economic theory may not completely specify the functional form of a model ‘Engel’s Law’ predicts a diminishing proportion of income will be spent on food as income rises This is consistent with the linear regression FOOD = α + β INCOME α > 0, β < 1 and the log-linear regression log(FOOD) = φ + γ log(INCOME) γ <1 We need statistical techniques to chose between these models
Background image of page 2
3
Background image of page 3

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

View Full DocumentRight Arrow Icon
Image of page 4
This is the end of the preview. Sign up to access the rest of the document.

This note was uploaded on 03/07/2012 for the course ECON 201 taught by Professor Cowell during the Spring '10 term at LSE.

Page1 / 7

Lecture 9 - Introduction to Diagnostic Testing In a...

This preview shows document pages 1 - 4. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online