About mike2468

mike2468 · ‎05-03-2017

Dynamic text is a great feature and very easy to implement. Very happy with 7.4 so far.

mike2468 · ‎04-26-2017

Is there an install guide or any other documentation that will assist with the 7.3 to 7.4 update process?

mike2468 · ‎01-24-2017

There are a few ways you could go about it. If you can get to the code that's doing the import, you can create the format as part of the import process. If you don't mind getting your hands dirty, you could do that to the autoload routine (it's just SAS code that runs on a schedule), but this is prone to be overwritten if/when SAS ever updates that code in a patch. I would probably move the data file out of the autoload directory on the server and write a custom program to load it (and create the format) that runs however frequently you need. Actually creating the format should not be too difficult. Just select out the distinct values of your column that contains the "Jul 13- Jul 19” type values, then chop up those strings and convert them to your "W01" values. Once you have a dataset that has the mapping for your format, you can use a "proc format cntlin" to convert that dataset into a format. I have some code around here somewhere that I can share if it will be helpful. You could also just run your format-creating code every 15 minutes or so, and have it fail nicely if the data set is not loaded (you can determine this from the _T_TABLEMEMORY table on your LASR server). You might even come up with some way to test to see if your format is outdated, rather than just rebuilding it every time...

mike2468 · ‎01-24-2017

How are you loading the data at present -- is it autoloaded from the Excel file?

mike2468 · ‎01-24-2017

You could procedurally generate your format based on the actual data.

mike2468 · ‎12-09-2016

I believe that's a list table with a big typeface and all the chrome set to white.

mike2468 · ‎12-06-2016

A decision tree would be a good starting point if you do not have Visual Statistics.

mike2468 · ‎11-29-2016

Thanks Sam. Happy to assist in any way I can.

mike2468 · ‎11-29-2016

It doesn't really matter...there are only actually five dates (one per year) in the entire dataset. So whether I aggregate it by quarter or year makes no difference. The reasons for this are complex but boil down to the fact that we work on a 52- or sometimes 53-week calendar. I have to filter my extract based on the actual comparison dates and then assign a synthetic date to each year of data...it happens to be September 1 in my report, but that could be anything. So the current set of data was extracted on 11/28/16, but I have to be able to calculate the change over the data from 11/30/15. In the extract I give the 11/28/16 data the "count date" of 9/1/16, and the 11/30/15 date the "count date" of 9/1/15. It is an ugly hack. If SAS could calculate change across variables that aren't dates, this would be a lot easier. This is a major shortcoming in the product IMO...Tableau and even Excel can do this with a couple of clicks. But I digress. "Year, Quarter" would not work (Fall 2016 has specific meaning for my consumers, and 2016, Q3 is not the same). At a push I could just use "Year" for the date format and specify elsewhere in the report that it is for fall. But this is not really satisfactory either. I am not surprised that it works on an earlier build. It worked on my instance until I updated it last week. I think that it may have been introduced with an update between about March/April (when I last updated) and hotfix 5. What version are you testing on? I could probably gin up a simple dataset and instructions to reproduce the issue I'm seeing if that will help. If someone else can verify that it fails on another install of 7.3 hotfix 5 and it still works on yours, that would support my suspicion that it may have been a regression introduced in the last six months or so.

mike2468 · ‎11-29-2016

Sure. I have attached three: one with the crosstab with the date variable (which works fine), one that shows the definition of the custom category based on the date variable, and one that shows the crosstab using the custom category instead of the date variable (which doesn't work). According to Help->About in the Designer I am on 7.3 Hotfix 5 with a build date of Oct 3. The specific formula for the "Change" measure is this: ParallelPeriod(_Sum_, 'Measure'n, 'my date variable'n, _ByYear_, _ByYear_, 0, _Full_, {Date}) - ParallelPeriod(_Sum_, 'Measure'n, 'my date variable'n, _ByYear_, _ByYear_, -1, _Full_, {Date}) and the % change is just the change divided by the last-year count. It does not seem to matter what format my date variable has. Thanks for the response. 🙂

mike2468 · ‎11-23-2016

I have a number of crosstab reports that use the ParallelPeriod functions to provide year-on-year change for measures. For example, I have a bunch of values in my datasource that are as-of 11/21/16, and a bunch more that are as of 11/23/15, and more as of 11/24/14, and so on. (We work on a 52-week calendar, which is why those dates shift about...they are all Mondays.) I need the change in the count of 2016 entries over the count of 2015 entries, and the change in 15 over 14, and so on. The ParallelPeriod function does this fine -- as long as I put that date variable in the rows or columns of the crosstab. The calculations are of this form: ParallelPeriod(_Count_, 'row unique identifier'n, 'Date of count'n, _ByYear_, _ByYear_, -1, _Full_, {Date}) However, I can't surface my dates to the report consumer (this is a business requirement). I have to use a descriptive name instead. So I need my 2016 entries to have the heading "Fall 2016", and the 2015 entries to have the heading "Fall 2015," and so on. So I create a custom category based on the as-of date. Everything with the date of e.g. "11/21/16" gets mapped to "Fall 2016," and the same with the other dates. Then I use that custom category in the crosstab instead of the date. This used to work fine. This week I applied about 6 months worth of patches (including the VA patch V90005) to my VA/BI stack, and it stopped working. Now all those ParallelPeriod calculations return "missing," and I get the exclamation-point in the bottom-right corner of the crosstab. The mouseover of the exclamation point says: "A mixture of date references caused missing values to be returned for a period calculation. If possible, consider creating the referenced dates by duplicating them from a common data item and adjusting properties of the duplicates such as the format." I can switch the custom category for the as-of date that it's based on, and the parallel calculations work fine. Duplicating the date and changing the fomat and other properties of the duplicate as suggested in the mouseover text does not help. Is this a bug -- perhaps something introduced in the latest round of patches? Any suggestions for a workaround?

mike2468 · ‎08-29-2016

I wrote a macro to do something similar. I wanted to check a table's load status and, if it is unloaded, fire a stored procedure that loads it. It's for reloading critical tables after a reboot, but you could adapt it for your needs. It uses the _T_TABLEMEMORY set in the LASR library. Instead of checking a specific table you will want to filter down the loadedtables dataset and then iterate through the rows to fire your PROC CONTENTS. data loadedtables; set VALIBLA._T_TABLEMEMORY; run; %macro ifunloaded(tablename, thescript); proc sql noprint; select count(*) into :OBSCOUNT from loadedtables where tablename = upper(&tablename) and uncompressedsize > 0; %if &OBscount = 0 %then %do; %put "&tablename is unloaded, we need to reload it"; proc stp program=&thescript; run; put "Table &tablename was unloaded and reload was attempted."; %end; %else %do; %put "&tablename is already loaded, leave it alone"; %end; %mend ifunloaded

mike2468 · ‎08-23-2016

This doesn't work (at least in 7.3). Checking "Required" automatically selects the top value in the dropdown. Even if it did, it would only work once. The report will always reopen with the last viewed date selected and not the most recent. It is confusing for users who may not realize they are seeing stale data. The simplest workaround I have found is to have one report for current data and one for historic. The current-data report does not have the date dropbox and gets a filtered view of the table that only includes the latest date. You could also add a field to your table that flags the most recent date and filter on that in the VA report -- have a prominent "current/final" switch tied to a parameter that applies a datasource filter, or break the report into two sections. I don't think you can use max(Date) in VA for this -- I think you will have to build it in your LASR table at extract time. I don't find any of these alternatives very satisfactory and am posting in the hope someone has a better idea.

Online Status	Offline
Date Last Visited	‎10-10-2017 06:27 PM

Re: SAS VA 7.4 release

Re: SAS VA 7.4 release

Re: stymied on line chart

Re: stymied on line chart

Re: stymied on line chart

Re: Create KPI-number without a meter

Re: Which is the best model to analyze Student Retention data?

Re: Possible bug with ParallelPeriods and custom categories on date va...

Re: Possible bug with ParallelPeriods and custom categories on date va...

Re: Possible bug with ParallelPeriods and custom categories on date va...

Re: SAS VA 7.4 release

Re: SAS VA 7.4 release

Re: SAS VA 7.4 release

Re: SAS VA 7.4 release

Re: stymied on line chart

Re: stymied on line chart

Re: stymied on line chart

Re: Create KPI-number without a meter

Re: Which is the best model to analyze Student Retention data?

Re: Possible bug with ParallelPeriods and custom categories on date va...

Re: Possible bug with ParallelPeriods and custom categories on date va...

Re: Possible bug with ParallelPeriods and custom categories on date va...

Possible bug with ParallelPeriods and custom categories on date variab...

Re: Programmatically check LASR Table Status

Re: VA 7.1 - Select the most recent date in a Dropdown list