- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
How the statistical procedures in SAS, such as proc means, proc model , to quote only a few, to read data from the SAS datasets? , read the all observations once a time to calculate the statistics, such as sample mean, or read one observation each time into PDV as DATA STEP and calculate the statistics each time? I can 't find the answer in SAS document and ask help.
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
It's not in the documentation. It's proprietary information. Most of it is a black box that we as users neither observe nor control. To a very small extent, some procedures (such as SQL) contain options that influence how observations are read. But that's all that is available.
When you mention the PDV, you are talking about a DATA step. There is no evidence that I know of that talks about how PROC steps handle this.
Because of the behavior of PROC MEANS (and other summary procedures), we can infer that SAS computes some statistics at a detailed level. For example, it computes SUM, N, and sum of squared values at a detailed level. Then as needed, it computes final statistics. (Use SUM and N to get the MEAN, use SUM, N, and sum of squared values to get the standard deviation. Aggregate the building blocks to a higher level in order to compute statistics at a higher level.)
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
Some statistical analysis procedures, such as Proc Model, Proc SSM, PROC MCMC have their programming statements, most of which are the same as those in DATA step. There is even a PDV option in OUTPUT statement of PROC SSM. Put things together , I infer that SAS PROCs read and deal with the observations in SAS data sets in the similar way to that in DATA STEP.