Yes, indeed!  A dataset for various code combination is something we're looking into as well.  The combinations we need to cover are quite extensive!   Some times its as simple as CC1 = A & B.  That is, if a patient has A and B the she meets criteria for CC1.  Slightly more complicated CC2 = A & IN (B, C).  Other times all three are required and it's CC3 = A & B & C.  And then you get two but not the third for CC4 = A & B & (NOT C).  And so on, and so on, and so on with about 200 base codes combining every which way two, three, four, ... fifteen at a time.   Thank you for your input!   ... View more

Benchmarking results are in: 1. My original code ran just a tad over two hours 2. My new and improved code without vValueX() or sorting improved significantly to 1.5 hours. 3. PROC IML method recommended above blew them both out of the water completing the task in 24.5 minutes!   Thanks for playing! ... View more

Thanks for this Danny.  Do you have a version for mixed models with random effects?  What you have here works wonderfully for marginal predicted probabilities.  How about conditional ones that include the random intercept for each level 2?   Thanks, Haris ... View more

Any input on MISSING() function?  Weird, isn't it?  It's fully documented which SAS functions can get sent to the DB for execution. Passing SAS Functions to ODBC     As to how I knew that SAS was importing the data: the original querry would (1) take way too long to retrieve <50 rows of data and (2) the memory.... You don't need to derive such information. Option sastrace writes to the SAS log which SQL gets sent to the database. If you would have used it and looked at the log you would have seen that SAS function missing didn't get translated to MS SQL syntax and though had to execute on the SAS side. options sastrace=(,,,d) nostsuffix sastraceloc=saslog;   ... View more

Thank you very much! ... View more

Thank you Steve. I was not looking to deliberate between PL and Quadrature approaches but this is certainly a relevant topic. Thanks for sharing! ... View more

Perfect! Thanks. Found the same solution. Any idea why one needs to request partitioning to get these basic model fit statistics to output? Does not make sense to me that they won't show up without typing extra code for the PARTITION option and specifying this contrived (TEST=0) scenario to, basically, negate the PARTITION request. ... View more

Thanks for the recommendation regardgin WORK.  I am looking at Windows Resource Monitor and don't see anything of note on CPU, RAM, or Disk I/O.  Maybe up to 10-15% utilization.  Anything else you can recommend? ... View more

I don't think you can do that with the JSON engine.  You basically want to read an array of JSON elements into a single character variable.     If you write your own program to parse the json file using INPUT, you should be able to do it.   Or maybe you could pre-process the JSON file to convert the arrays into string values, i.e. change:   "proc": [ 2800, 2980, 4090, 4170 ], to: "proc": "[ 2800, 2980, 4090, 4170 ]", ... View more

Does not look like the format of the MAP files that JSON libname engine uses. You could try to convert that into a definition of a dataset and/or all the way into a MAP so that you are ready when the real file comes.  You should ask for an example with a few dummy records you can use to test it.   You could try converting by hand.  Or if it is really large perhaps you could convert that JSON into some SAS datasets and then use a program to use that write a MAP file for SAS.  Or perhaps just regular SAS code to define the dataset(s) you will need to create from the JSON text.   For example it looks like there is a variable named DXOUT_IDENTIFIER that should be defined as length $255. ... View more

If you are IO bound then it is your SAS WORK space that will be the bottleneck. Which drive does SAS WORK go to now? This will tell you where: proc options option = work; run;   Typically you point to a different SAS WORK space in your SAS config file but it can be done as an option in the SAS batch command ... View more

Propensity score of 1 means that the probability of the target treatment is 100% for a company based on your model and the values of the covariates. ... View more

BTW, if helpful, single models took 8.5 hours to run.  When running three at the same time, the execution time increased to 9.5 hours for each but total start-to-finish time was clearly saved in concert. ... View more

I often want to temporarily keep a report available that would be overwritten by running a portion of the same code file. The quickest workaround for me to keep the report available is, while on the results tab, right click the report name in the left side index and "share" it with another application. Sharing it with "Send to Windows default" is probably the best option for most users and if you haven't changed the default report style from HTML (in the Tools > Options > Results menu) then that should send it to a new tab in your web browser. ... View more