Regression.txt

 March 182, 2003

 http://www.ruf.rice.edu/~lane/rvls.html
 We don't have CRSP+COMPUSTAT integrated, or CRSP Mutual funds. 


g604, IUg604g604

 For laptop: FUNCTION F7 cycles through the RGB stuf-- both screens,

XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

 FIRST HOUR: Simple  Regression

Introduction: 
  Good morning. 

You always understand your next-to-last econometrics class. 


 Introduction. Overhead with 2-day plan.

Start with Regression to the mean. Orley.

1. Sports Illustrated Jinx.

2. Air force pilots--use criticism, not praise.





   The Law Firm Associates Example.

    The Problem and the Data.

    Summary Statistics

 Graphing Data on IQ, HOURS



  Quantifying the relationship: move to Rubinfeld on correlation.
 Note that slope and correlation are not the same thing.

 Correlation matrix for Associates




 Simple regression, in Rubinfeld.  Drawing a line through the middle.

 The idea of least squares.  Least absolute deviations.

 The equation of Rubinfeld.

Simple regression for Associates, IQ and Pay.

The importance of the intercept. If you assume the true relationship has
(0,0) (that is--  constraining intercept to be zero), that makes a
difference.




     At this point, show them how to use Excel to:

 1. Graph the data

 2. Find correlation matrix

 3. Do the simple regression, with and without intercept




  What about eye-balling the data instead?

Applet for   Eye-drawing regression line. Simple correlation.



XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX





MULTIPLE REGRESSION

Then do a multiple regression.

 Residuals on individuals.


Discuss t-tests

COnfidence Interval.  Tests, not regression.   

 



 

Discuss R2.

Small Effect size.  R2. Not regression.


Omitted Variables, correlated with sex-- suppose  the women are more
quarrelsome, and don't get along with the other associates or
paralegals.







 Then do with less data.

Then do, just doubling each data point.



Nonlinearity of seniority.

Use applets for: Transformations. Log, Square root, etc. DO ON WEB.

XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

SECOND HOUR: INFO IN DATA.  LIMITED DEPENDENT VARIABLE. DATA PROBLEMS.

 Components of r. Simple correlation.  Dispersion of data useful.
Restricted range: Strength, SAT. Simple correlation. R2. Dispersion of
data useful.

  Judges in Japan. Institutions from Chapter 1.

  Do the judges example from Ramseyer and Rasmusen.

Dependent variable a dummy variable

XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

 
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX



 

XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX