|
|
Dear list members,
I have a simple situation with data as below, with missings at variable y.
data list free / year,y.
begin data.
1990,00 6,00
1995,00 5,12
1998,00 10,00
1999,00 13,00
2005,00 12,00
1991,00 999,00
1992,00 999,00
1993,00 999,00
1994,00 999,00
1996,00 999,00
1997,00 999,00
2000,00 999,00
2001,00 999,00
2002,00 999,00
2003,00 999,00
2004,00 999,00
End data.
missing values y(999).
When running a regression with:
reg /dep=y /method=enter x /save=pred(pred).
The result y = -1001.210 + 0.506*x can be found.
I also executed a MVA (no error term added fixed tolerance and flimit) with
predicted unstandarised value returned by
regression, with syntax:
mva y x /regression y with x (tolerance=.001 flimit=4 addtype=none
outfile='c:\test.sav' ) .
Finally I wanted to find out the equation of the imputed values, so I did:
get file='c:\test.sav'.
compute volg=$casenum.
formats volg(f2).
execute.
temporary.
select if (volg>5).
reg /dep=y /method=enter x.
The result is an equation for the imputed values of: y = -1341.556 + .676*x.
My question is: why is this result different from the result of the above
mentioned regression analysis.
Probably MVA does not use OLS?
Any help is highly appreciated!
Pieter van Groenestijn
University of Nijmegen
The Netherlands
1,00 999,00
1992,00 999,00
1993,00 999,00
1994,00 999,00
1996,00 999,00
1997,00 999,00
2000,00 999,00
2001,00 999,00
2002,00 999,00
2003,00 999,00
2004,00 999,00
End data.
I have different values from regression and MVA. Anyone know why this values
A similair question is posted by Jurek Madej, May 2005. Unfortunately he
did not receive an answer.
|
Locked
