# Help with effects statements

#### trailblazer

##### New Member
I am trying to write effects statements and the online materials are difficult to understand.

I am working with the below linear regression (proc regress) model in SAS callable SUDAAN and want to determine the beta coefficient for the effect of violence. I would like an overall estimate, and also one for men (sex=1) and women (sex=2). Can someone please help me? I think the effects statements are the same as estimate statements in other SAS procedures. Thanks!

Model
fruitveg = violence + sex + age + race + education + foodst + tv+ bmi + soda + internet + active + violence*sex

Dependent variable
fruitveg= Servings of Fruits and Vegetables per day=continuous

Independent Variables
violence= 1=yes, 2=no
sex, 1=male, 2=female
age=continuous age
race= 1=Black, 2=Hispanic, 3=Asian, 4=Other, 5=White
education 1=Less than HS, 2= Completed HS Degree, 3=Completed College Degree
foodst= receipt of food stamps 1=yes, 2=no
tv= number of tv hours watched per day (continuous)
bmi, 1=underweight, 2=normal weight, 3=overweight, 4=obese
soda, number of soda drinks per day (continuous)
internet, number of soda drinks per day (continuous)
active (physically active) 1=yes, 2=no

#### hlsmith

##### Less is more. Stay pure. Stay poor.
What does your current coded attempt look like?

#### trailblazer

##### New Member
I haven't attempted code because there is no way for me to know if it is yielded the correct output. I am generally confused by effect/contrast statements, is there anyone here who is knowledgeable?

#### Buckeye

##### Active Member
From my understanding if you have the estimated beta coefficients of a regression model the overall anova effect is twice that value. Put another way, if you have an arbitrary effect the estimated coefficient is (effect)/2 I hope that's what you're asking. This assumes your predictors are coded from -1 to 1. The estimate statement is a way to construct a contrast. Now that I think about it, I believe my explanation only applies to 2^k factorial designs.

Last edited: