BPPANELTESTS Procedure

@BPPANELTESTS does the Breusch-Pagan LM test for random effects (and several variants on it), from Breusch & Pagan(1980). This is a test based upon the OLS residuals, and needs to be (immediately) preceded by that regression. It does the original LM version, the signed version from Honda and a signed version with small-sample corrections (the SLM or standardized LM in the output).

@BPPanelTests( options ) u

Parameters

u	series of residuals from the OLS regression (LINREG or PREG(METHOD=POOLED)). It should not be fixed or random effects.

Options

EFFECTS=[INDIV]/TIME/BOTH

Panel effects to test for.

TITLE="title for report" ["Breusch-Pagan Tests for Random Effects"]

Example

open data tablef8-1.csv

calendar(panelobs=7,a) 1976

data(format=prn,org=columns) 1//1976:01 595//1982:01 exper wks occ ind $

south smsa ms fem union ed blk lwage

set expersq = exper^2

preg(method=pooled) lwage

# exper expersq wks occ ind south smsa ms union constant

@BPPanelTests(title="LM Tests for Common Effects") %resids

Sample Output

This shows the three variants, with the test statistics and the associated significance levels. The Breusch-Pagan is asymptotically a chi-square with 1 degree of freedom, the other two are N(0,1) with one-sided tests.

LM Tests for Common Effects

Breusch-Pagan 39.7272

(0.0000)

Honda 6.3029

(0.0000)

SLM 10.4451

(0.0000)