The RATS Software Forum

Hi,

When I run these instructions @lsdvc(met=ah) retuns
#MO TU WE TH FR


I get the following output.
retuns are daily stock retuns, while MO,TU, WE, TH and FR are dummies for days of the week.
Why do I get the error? What did I do wrong? Can i still use the output?

Thanks,
sulong

The Error Occurred At Location 1228, Line 87 of LSDVC
## REG10. First 6 Instruments Are Linearly Dependent Over Regression Range

Linear Regression - Estimation by Anderson-Hsiao
Dependent Variable RETUNS
Panel(22) of Daily(5) Data From 1//1900:01:02 To 66//1900:01:30
Usable Observations 1386
Degrees of Freedom 1380
Skipped/Missing (from 1451) 65
Mean of Dependent Variable 0.4462584441
Std Error of Dependent Variable 3.1202276003
Standard Error of Estimate 3.1250529909
Sum of Squared Residuals 13477.019550
Durbin-Watson Statistic 1.9575

Variable Coeff Std Error T-Stat Signif
************************************************************************************
1. RETUNS{1} -0.000257156 0.027464333 -0.00936 0.99253064
2. MO 0.371827816 0.172548844 2.15491 0.03134116
3. TU 0.282525836 0.210761775 1.34050 0.18030398
4. WE 0.240602521 0.210760562 1.14159 0.25382180
5. TH 0.506416548 0.172155503 2.94162 0.00331944
6. FR 0.000000000 0.000000000 0.00000 0.00000000

You can't use a full set of weekday dummies since LSDVC includes (by construction) a full set of individual dummies. If you drop Friday, you'll get exactly the same results without the warning as RATS simply drops the redundant dummy (thus the zero coefficient, zero standard error).

You are right, no more error!
But, what about a constant Tom?
Should I include that since I take out one dummy variable?


appreciate your help
sulong

No. The implied full set of individual dummies are collinear with the constant, so you don't want to include that.

okay,
Thank you!