Using SUDAAN in SAS OnDemand for Academics

[_maldini_]

Does SAS OnDemand for Academics support SUDAAN?

The SUDAAN support site states: "SAS users will want to get the SAS-Callable version of SUDAAN, which is installed as an add-on to SAS on your Windows or Sun/Solaris computer system."

I'm using SAS OnDemand for Academics on a Mac. 

[SASKiwi]

SAS UE has been discontinued and replaced by SAS OnDemand for Academics. SODA runs on linux so doesn't appear to be compatible with SUDAAN. In any case you would need to ask SAS regarding any software additions as users can't change that.

[_maldini_]

My apologies. I'm running OnDemand for Academics. I changed the original post.

Thanks for your response.

[Kurt_Bremser]

Since SUDAAN does not support the platform on which SODA runs, SAS could not even install it, even if they wanted to (it would have to be installed on the SAS server, not on the client).

You can only try to use the stand-alone version of SUDAAN in a Windows partition, if your Mac still supports that.

Or get a Windows PC.

[ballardw]

You may consider discussing exactly which feature of SUDAAN you need that is not available in the SAS Surveyprocs to see if someone has a work around.

For some basic analysis years ago I did compare output from SUDAAN and the SAS survey equivalents and noticed some differences in point estimates or confidence limits. Typically the differences occurred such as 53.4359 vs 53.4358 or smaller differences. Since for my work I seldom going to report anything past that first decimal place it wasn't considered worth the extra expense for a SUDAAN license, and the associate headaches like no ODS output and SUDAAN failing when SAS was upgraded.

[_maldini_]

I'm trying to perform a trend analysis (e.g., hypertension prevalence over time) using weighted survey data from NHANES. I posted about this topic in this forum previously, and then found a page on the CDC website that seemed to indicate this could only be done in SUDAAN (also see image. "U" is SUDAAN. "S" is SAS). 

The differences you describe in your post are not meaningful to me either. One or two decimal places is all I need.

Thanks for your thoughts here.

[ballardw]

You want to read the introductions carefully. That graphic is based on "Below is a listing of sample code demonstrating some statistical analysis topics in select NCHS publications using NHANES." (Emphasis added)

So that is just some approaches that the website can provide example code to duplicate what the publications reported. It does not in any way indicate it is the only way to perform analysis.

Some of the trend analysis done in the SUDAAN code using SUDAAN Proc Ratio, others are chi-square or regressions which could be done with SAS Proc Surveyfreq with the CHISQ option on Tables statement or Proc Surveyreg.

I don't have any easy to find comparisons of Proc Ratio at the moment. So look in the code examples for the specific trends you are interested in, or at least similar.

[_maldini_]

Thanks for the clarification on the CDC website. 

<Some of the trend analysis done in the SUDAAN code using SUDAAN Proc Ratio, others are chi-square or regressions which could be done with SAS Proc Surveyfreq with the CHISQ option on Tables statement or Proc Surveyreg.>

Objective #1: To test for trends in categorical variables (e.g., hypertension (y/n), cigarette smoking (never/past/current), etc.) over time, in this case, years, after adjusting for sex and age. I want to report the age and sex adjusted proportions and 95% CIs.

Q1: Are you saying that adding the CHISQ option to the Tables statement in Proc Surveyfreq will test for a trend (assuming ordinal values in the row or column variable of interest. In this case "years")?

If so, this would provide the p value for the trend analysis, but not the adjusted proportions and 95% CIs, correct?

Could I use Proc Surveylogistic (somehow) to derive the adjusted proportions and 95% CIs?

Objective #2: To test for trends in continuous variables (e.g., cigarette-pack years, BMI, etc.) over time, in this case, years, after adjusting for sex and age. I want to report the age and sex adjusted means and 95% CIs.

Q2: Can I use Proc Surveyreg to analyze a trend in continuous measures (e.g., cigarette-pack years over time)? If the model includes age and sex (in addition to the variable for time (i.e., years)), how would I request the adjusted means? 

This is difficult to explain. I hope I'm making sense. Thanks for any help.