MEANGROUP Procedure

@MEANGROUP does pooled mean group estimation for panel data. Pooled mean group estimates the coefficient vector of the regression by an equally-weighted average of the individual estimates. @SWAMY is a similar procedure, but does the Swamy random coefficients GLS estimator.

@MeanGroup( options ) depvar start end

# list of regressors (in regression format)

Parameters

depvar	dependent variable
start, end	range for regression. By default, the maximum range permitted by all variables involved in the regression.

Options

SMPL=standard SMPL option [none]

[PRINT]/NOPRINT

TITLE="title for output" ["Pooled Mean Group"]

DEFINE=equation to define [none]

Variables Defined

%BETA	estimates of mean of coefficient process (VECTOR)
%XX	estimates of covariance matrix of mean coefficient (SYMMETRIC)
%%IBETAS	individual coefficients (VECT[VECTOR]). %%IBETAS(i) is the coefficient vector for individual i.
%%IXX	individual covariance matrices (VECT[SYMMETRIC]). %%IXX(i) as the covariance matrix for individual i

Example

* Based upon Greene, Econometric Analysis, 7th Edition

* Example 11.19 from p. 418-419

open data produc_fix.prn

calendar(panelobs=17,a) 1970:1

data(format=prn,org=columns,left=3) 1//1970:01 48//1986:01 yr p_cap hwy water util pc gsp emp unemp

* Log all variables except unemp

dofor s = gsp p_cap hwy water util pc emp

log s

end dofor

linreg gsp

# constant pc hwy water util emp unemp

@meangroup gsp

# constant pc hwy water util emp unemp

Sample Output

This shows the output from the LINREG (OLS) and @MEANGROUP. Both are weighted averages of the individual estimates, but LINREG gives "precision-weighted" averages (giving higher weights to the individuals with high X'X) and @MEANGROUP uses simple averages.

Linear Regression - Estimation by Least Squares

Dependent Variable GSP

Panel(17) of Annual Data From 1//1970:01 To 48//1986:01

Usable Observations 816

Degrees of Freedom 809

Centered R^2 0.9930534

R-Bar^2 0.9930018

Uncentered R^2 0.9999351

Mean of Dependent Variable 10.508849637

Std Error of Dependent Variable 1.021131864

Standard Error of Estimate 0.085422808

Sum of Squared Residuals 5.9033184420

Regression F(6,809) 19275.0232

Significance Level of F 0.0000000

Log Likelihood 853.1372

Durbin-Watson Statistic 0.1877

Variable Coeff Std Error T-Stat Signif

************************************************************************************

1. Constant 1.926004375 0.052503182 36.68357 0.00000000

2. PC 0.312023086 0.011087500 28.14188 0.00000000

3. HWY 0.058881719 0.015411448 3.82065 0.00014330

4. WATER 0.118580557 0.012356570 9.59656 0.00000000

5. UTIL 0.008555123 0.012354029 0.69250 0.48882423

6. EMP 0.549695456 0.015536879 35.38004 0.00000000

7. UNEMP -0.007270503 0.001383632 -5.25465 0.00000019

Linear Regression - Estimation by Pooled Mean Group

Dependent Variable GSP

Panel(17) of Annual Data From 1//1970:01 To 48//1986:01

Usable Observations 816

Degrees of Freedom 809

Mean of Dependent Variable 10.508849637

Std Error of Dependent Variable 1.021131864

Standard Error of Estimate 0.392344723

Sum of Squared Residuals 124.53291476

Durbin-Watson Statistic 0.0873

Variable Coeff Std Error T-Stat Signif

************************************************************************************

1. Constant 0.964589401 0.639786148 1.50767 0.13163780

2. PC 0.027555729 0.036106378 0.76318 0.44535501

3. HWY 0.086149959 0.102872837 0.83744 0.40234458

4. WATER 0.053177878 0.044939890 1.18331 0.23668581

5. UTIL 0.066369134 0.078099686 0.84980 0.39543615

6. EMP 1.041287097 0.067057858 15.52819 0.00000000

7. UNEMP -0.002475663 0.001438341 -1.72119 0.08521578