lect11_103_2010_Compatibility_Mode_

lect11_103_2010_Compatibility_Mode_ - 5/12/2010 A common...

Info iconThis preview shows pages 1–3. Sign up to view the full content.

View Full Document Right Arrow Icon
5/12/2010 1 Multiple Regression in Practice - Intro • A common question in practice is which X variables to include in a multiple regression. • Issue 1: Coefficient estimates will always change at least a little bit (and sometimes a lot) when we add, subtract, or change subsets of the regressors (i.e. the X ’s) . • Issue 2: Because of Issue 1, an important decision is what variables to include as X ’s in a regression. The main cost-benefit tradeoff here is: – 1) Not including a variable can lead to OVB – 2) Including too many variables can lead to Imperfect Multicollinearity , i.e. when the X variables are highly (but not perfectly) correlated with each other. Imperfect Multicollinearity can result in high standard errors, i.e. imprecise estimates. For example, with very large standard errors, one gets very wide and uninformative confidence intervals. General Strategies - I • Suppose you are primarily interested in quantifying how one specific regressor ( X 1i ) affects Y i . • You have a set of possible additional regressor variables X 2i , X 3i ,…., X Ki . Which should you include? • First, use your intuition to come up with a “baseline” specification, i.e. an initial set of regressors. • In this baseline specification:
Background image of page 1

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

View Full DocumentRight Arrow Icon
5/12/2010 2 General Strategies - II • 1) Include (as X’ s) variables that you think are likely to be important determinants of Y i . • 2) Do not include (as X’ s) variables that measure the same thing as Y i , or that are “caused” by Y i . This can cause “simultaneity bias” (see pp 324-325 in text). Examples: – Suppose that test scores are the average of a math score and a verbal score, i.e. testscore = 0.5*mtestscore + 0.5*vtestscore. You should not include mtestscore and vtestscore as regressor variables. What would happen? – Suppose you also measure 9 th grade testscores. Since these are in some sense “caused” by 5 th grade testscores, you should not include this as a regressor variable. So include regressors that you think cause Y i , do not include regressors that you think are caused by Y i . General Strategies - III • 3) In constructing the baseline specification, if you have two (or more) X ’s that are similar measures of the same effect, you often want to start by only including one (or a few) of these X ’s • Example: In our Testscore-STR study, suppose we also want to
Background image of page 2
Image of page 3
This is the end of the preview. Sign up to access the rest of the document.

This note was uploaded on 06/17/2010 for the course ECON 103 taught by Professor Sandrablack during the Spring '07 term at UCLA.

Page1 / 8

lect11_103_2010_Compatibility_Mode_ - 5/12/2010 A common...

This preview shows document pages 1 - 3. Sign up to view the full document.

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