Chow Test

[smukherjee01]

I have a sample which I need to break into 3 groups. I have done Chow Test for 2 groups, but never with 3 groups in RATS.

Is there something wrong if I do the following in RATS?

Code: Select all

OPEN DATA "EXAMPLE1.xlsx"
DATA(FORMAT=XLSX,ORG=COLUMNS) 1 167 COUNT FOR  FREV FER OFM 
SET TPFE = FER *10
SET TPFR = FREV*10
SET BPOF = (OFM*50000)/10
LINREG TPFE
#CONSTANT TPFR BPOF
COMPUTE RSSSMALL=%RSS , NDFSMALL=%NDF
DISPLAY RSSSMALL
DISPLAY NDFSMALL                                              


OPEN DATA "EXAMPLE2.xlsx"
DATA(FORMAT=XLSX,ORG=COLUMNS) 1 220 COUNT FOR  FREV FER OFM 
SET TPFE = FER *10
SET TPFR = FREV*10
SET BPOF = (OFM*50000)/10
LINREG(SMPL=COUNT<387) TPFE
#CONSTANT TPFR BPOF
COMPUTE RSSMED=%RSS , NDFMED=%NDF
DISPLAY RSSMED
DISPLAY NDFMED



OPEN DATA "EXAMPLE3.xlsx"
DATA(FORMAT=XLSX,ORG=COLUMNS) 1 115 COUNT FOR  FREV FER OFM 
SET TPFE = FER *10
SET TPFR = FREV*10
SET BPOF = (OFM*50000)/10
LINREG TPFE
#CONSTANT TPFR BPOF
COMPUTE RSSLARGE=%RSS , NDFLARGE=%NDF
DISPLAY RSSLARGE
DISPLAY NDFLARGE




OPEN DATA "EXAMPLE4.xlsx"
DATA(FORMAT=XLSX,ORG=COLUMNS) 1 501 COUNT FOR  FREV FER OFM OFQ2M OFQ1M 
SET TPFE = FER *10
SET TPFR = FREV*10
SET BPOF = (OFM*50000)/10
LINREG TPFE
#CONSTANT TPFR BPOF
COMPUTE RSSPOOL=%RSS
DISPLAY RSSPOOL


COMPUTE RSSUNR = RSSSMALL+RSSMED+RSSLARGE
COMPUTE NDFUNR = NDFSMALL+NDFMED+NDFLARGE
COMPUTE FSTAT = ( (RSSPOOL-RSSUNR)/3 )/(RSSUNR/NDFUNR)

CDF FTEST FSTAT 3 NDFUNR

[TomDoan]

You have your degrees of freedom wrong---it would be 6 (9 total coefficients in the 3 way split vs 3 in the pooled sample).

The setup seems a bit odd though. If the Example4 data set has the whole sample, can't you get the subsamples out of that without creating three extra data files?