One coefficient is different, rest same....

[jb1983]

Hi

I am a newbie to SAS so please bear with me.

I am trying to replicate some results from regressions that I have, these were supposedly run in SAS but thats not for sure. I used the SAS code to run the model and got ALL standard deviations and all but one coefficients corrrect in the proc reg command.

The coefficient that is incorrect would be correct if I do 1-beta_1.

I suspect the reason is because of how SAS runs regressions is different in some way and I would like someone's help in transforming the reg input so get the same out as I have from outside. I need this because of many other regressions and because I need ot run ridge regressions which are producing very different results.

Here is the code, the SAS output and the desired output:

proc reg data=Data
outest=b ridge=0 to 5 by 0.5;
StartingDate1950: model y_Log = y_Log_Ch_Lag12  xYoY_Lag12 z_Ch_Lag12 xYoY_Ch_Lag12 y_Log_Lag12;
weight Sqrtx;
where 195001<=Eom<=200512 and CountryID^=100;
run;
proc print data=b;
run;

SAS Output

Parameter Estimates Variable DF Parameter
Estimate Standard
Error t Value Pr > |t| Intercept 1 xYoY_Lag12 1 z_Ch_Lag12 1 xYoY_Ch_Lag12 1 y_Log_Ch_Lag12 1 y_Log_Lag12 1

-0.53073	0.01686	-31.49	<.0001
3.63783	0.11334	32.10	<.0001
4.43167	0.24145	18.35	<.0001
-1.35277	0.09557	-14.15	<.0001
-0.19969	0.01158	-17.24	<.0001
0.84583	0.00621	136.29	<.0001

Desired output

Intercept		-0.53073	0.016855
y_Log_Lag12		-0.15417	0.006206
xYoY_Lag12		3.637828	0.113342
z_Ch_Lag12		4.431668	0.241454
xYoY_Ch_Lag12		-1.35277	0.095572
y_Log_Ch_Lag12		-0.19969	0.011581

Thanks!!

[Reeza]

Is the one coefficient that's wrong an indicator variable, ie 0/1?

[jb1983]

No, it is a time series

[Rick_SAS]

Any chance that the original analysis was run with the NOINT option on the MODEL statement?

[jb1983]

Well, I am not sure what the exact code was so its possible.

But

when i run with noint now, it gives a completely different set of coefficient estimates.

Maybe Im not understanding what youre saying...

[jb1983 · Accepted solution]

Sorry, I should have said, the original regression output data has intercepts that I can match to mine as well.

So they have intercepts