BookmarkSubscribeRSS Feed
🔒 This topic is solved and locked. Need further help from the community? Please sign in and ask a new question.
saramanzella
Fluorite | Level 6

Hello everyone!

 

I'm using SAS Studio.

I'm applying the proc causalmed to some data for my thesis but I'm having some trouble understanding how it works:

  1. Can anyone tell me what does the "Scale" term at the end of every model table mean?
  2. The treatment I'm using is continuous so I was wondering how the procedure obtains the estimates of the effects in this case. I know the theory but I'm having some trouble understanding how the procedure actually calculates the effects.

I read the entire SAS Guide but didn't have any luck finding the answers.

I'm hoping someone else had the same questions I have and found an answer somewhere else.

 

Many thanks,

Sara

 

 

1 ACCEPTED SOLUTION

Accepted Solutions
MichaelL_SAS
SAS Employee

Good question, for a continuous treatment variable the natural direct effect (NDE) and natural indirect effect (NIE) estimates will depend on what levels are used for the treatment-to-control comparison. How the treatment and control levels are determined is discussed in the section "Default Settings of Treatment and Control Levels" in the documentation for PROC CAUSALMED

 

https://go.documentation.sas.com/?docsetId=statug&docsetTarget=statug_causalmed_details07.htm&docset...

 

For a continuous treatment, the default uses the mean of the treatment variable plus 0.5 as the treatment level, and the mean minus 0.5 as the control level. You can use the EVALUATE statement in PROC CAUSALMED to request the analysis at different levels.

For example, if you use the data set in the “Getting Started” example for PROC CAUSALMED the following code will produce three effect summary tables, one at the default comparison levels and two more for the comparisons specified in the EVALUATE statements.

 

proc causalmed data=Cognitive;
   model    CogPerform  = Encourage Motivation;
   mediator Motivation  = Encourage;
   covar FamSize SocStatus;
   evaluate 'Same as default' _treatment=34.916 _control=33.916;
   evaluate 'Different than default' _treatment=35.916 _control=32.916;
run;

After using PROC MEANS to get the mean of the treatment variable, I set the levels in the “Same as default” comparison so that it should match the default values. The effect summary for the “Different than default” comparison should show different estimates for the NDE and NIE than the other two tables. If you want to run this example, you can get the code needed to create the Cognitive data set by going to “Getting Started” section of the PROC CAUSALMED documentation online (HTML version) and clicking the link “code for this example” at the beginning of the section.   

 

https://go.documentation.sas.com/?docsetId=statug&docsetVersion=15.1&docsetTarget=statug_causalmed_g...

View solution in original post

3 REPLIES 3
SAS_Rob
SAS Employee

The scale parameter is the estimate for the dispersion parameter for a given distribution.  For the default normal distribution it would be an estimate of the standard deviation.

 

Which estimates are you referring to specifically?  If you mean either outcome model or median model estimates then the values are simply used in the design matrix directly (which would only have a single column rather than multiple columns of dummy variables when you have a CLASS variables) and then the estimates are computed using MLE.

saramanzella
Fluorite | Level 6

When I wrote estimates I meant the Natural Direct/Indirect Effects that the proc provides in the output. I know that the theoretical expressions to obtain those compare two possible values of the exposure variable. If this variable is a dummy then there's no problem, but if it's a continuous variable what values are compared? Or are those simply the estimates of the effects for a variation of 1 the exposure?

Thank you so much for the answer,
Sara

MichaelL_SAS
SAS Employee

Good question, for a continuous treatment variable the natural direct effect (NDE) and natural indirect effect (NIE) estimates will depend on what levels are used for the treatment-to-control comparison. How the treatment and control levels are determined is discussed in the section "Default Settings of Treatment and Control Levels" in the documentation for PROC CAUSALMED

 

https://go.documentation.sas.com/?docsetId=statug&docsetTarget=statug_causalmed_details07.htm&docset...

 

For a continuous treatment, the default uses the mean of the treatment variable plus 0.5 as the treatment level, and the mean minus 0.5 as the control level. You can use the EVALUATE statement in PROC CAUSALMED to request the analysis at different levels.

For example, if you use the data set in the “Getting Started” example for PROC CAUSALMED the following code will produce three effect summary tables, one at the default comparison levels and two more for the comparisons specified in the EVALUATE statements.

 

proc causalmed data=Cognitive;
   model    CogPerform  = Encourage Motivation;
   mediator Motivation  = Encourage;
   covar FamSize SocStatus;
   evaluate 'Same as default' _treatment=34.916 _control=33.916;
   evaluate 'Different than default' _treatment=35.916 _control=32.916;
run;

After using PROC MEANS to get the mean of the treatment variable, I set the levels in the “Same as default” comparison so that it should match the default values. The effect summary for the “Different than default” comparison should show different estimates for the NDE and NIE than the other two tables. If you want to run this example, you can get the code needed to create the Cognitive data set by going to “Getting Started” section of the PROC CAUSALMED documentation online (HTML version) and clicking the link “code for this example” at the beginning of the section.   

 

https://go.documentation.sas.com/?docsetId=statug&docsetVersion=15.1&docsetTarget=statug_causalmed_g...

SAS Innovate 2025: Register Now

Registration is now open for SAS Innovate 2025 , our biggest and most exciting global event of the year! Join us in Orlando, FL, May 6-9.
Sign up by Dec. 31 to get the 2024 rate of just $495.
Register now!

What is ANOVA?

ANOVA, or Analysis Of Variance, is used to compare the averages or means of two or more populations to better understand how they differ. Watch this tutorial for more.

Find more tutorials on the SAS Users YouTube channel.

Discussion stats
  • 3 replies
  • 2312 views
  • 5 likes
  • 3 in conversation