Programming the statistical procedures from SAS

Regression Model Using Dummy Variable

New Contributor
Posts: 2

Regression Model Using Dummy Variable



I am a SAS toddler and running a regression model using dummy variable for an event study. My model is


Realized Return = Expected Return + Dummy for event day (Day -10 to Day +10)


proc reg data=work.germanyindex;
model RR = ER DB10 DB09 DB08 DB07 DB06 DB05 DB04 DB03 DB02 DB01 DDAY DA01 DA02 DA03 DA04 DA05 DA06 DA07 DA08 DA09 DA10;


In the model, RR means realized return, and ER is expected return, and DB10 to DA10 is date dummy consist of 0 or 1. I have around 270 observations in my dataset.


However, if I run this regression, SAS says:



Note:Model is not full rank. Least-squares solutions for the parameters are not unique. Some statistics will be misleading. A reported DF of 0 or B means that the estimate is biased.


Note:The following parameters have been set to 0, since the variables are a linear combination of other variables as shown.


and I cannot get coefficient of ER variable nor std. error, t-stats, and prob. 


I do not know what can I do for this. Is it a data problem? or coding? I attached my dataset just in case. 


I will appreciate your help.

Super User
Posts: 18,529

Re: Regression Model Using Dummy Variable

Remember when coding dummy variables you include N-1 variables for N levels. 

This is because if you have 3 levels you only need two variables to uniquely identify all 3 scenarios.

New Contributor
Posts: 2

Re: Regression Model Using Dummy Variable

You mean number of dummy variables by levels? I have dummy variables for each 21 days, so I need to remove one of dummy variable?


Or, values of variable? DB10 - DA10 is consist of 1 and 0, but do I need to make blank for 0s?


Sorry for asking you additional question.

I am real beginner of SAS, so these question could be stupid. please understand.

Super User
Posts: 18,529

Re: Regression Model Using Dummy Variable

This isn't a SAS question, it's a statistics question. SAS is simply a program for applying statistical knowledge, you still need to understand the statistical methodology.


If you have a variable that can be expressed as a linear combination of other variables, it is not independent. So if you have 0/1 for every day, then you have overspecified the model and need to remove one of the day variables. 


A more detailed write up:



Ask a Question
Discussion stats
  • 3 replies
  • 2 in conversation