Winsteps
Sunday, July 31, 2005
Plot before you t-test!
Starting to catch up with the Paper reviews. One paper shows a plot and reports significance tests. The plot clearly says "There's a trend!" The t-tests say "No there isn't!" Reason: they are the wrong t-tests for the situation. Solution: Look at the plot before you t-test.
Dallas Workshop, May-June 2005
Gregory C. has emailed a couple of photographs. Wonderful to be reminded of the the great folks at the Workshops in Dallas! Left to right: Sandy P., Livia M., Gregory C.
 |  |
Saturday, July 30, 2005
Winsteps Rasch Explorations
2 months ago were the Winsteps and Facets workshops in Dallas. A month ago was the Winsteps/Facets workshop graciously hosted by the International Islamic University in Kuala Lumpur. 10 days ago was the Winsteps workshop at the University of Sydney. All were very busy and productive. It's good to meet with users at all levels of experience, and from so many backgrounds. You folks really keep me on my toes. And your suggestions for software enhancement are so obvious that they are excellent!!
It is really great to have others picking up the workshop load. Ken & Nick tell me their Winsteps Workshop in Chicago a few days ago really hummed. Nick writes: You will be pleased to learn that some of our participants were having a "riot" with WINSTEPS graphing procedures. At a certain point, they left the main group to investigate DIF issues.
And, after that, Carol and Lidia conducted a Facets workshop. Carol writes: This was a really fun group to work with, and they seemed to develop their own group spirit over the two days (and are interested in staying in touch with one another).
Meanwhile, the Winsteps Summer Sale was humming along. It was an experiment motivated by Henry Ford. Before his time, automobiles were an expensive toy for the rich and technically oriented. He perceived the car to be the means of transport for the masses. For that to happen, he needed to produce a higher quality and much more reliable product at a much lower price. Hence the production line, more robust components, etc. And the Model T success story followed. Now we all have cars. Winsteps is already, in many respects, higher quality, more reliable software. It is regularly sold at a lower price. But what if the price was halved - down to the price of shareware? That was the Summer Sale. Initial reports from Digibuy, our software sales agent, are favorable - but what will sales look like for the rest of the year?
Sometimes the simplest questions are the most profound. The simple question prompts some thought: We need to restrict the range of our logits output to plus and minus 3. How can we do this?
That was all there was in the email. Did they want to rescale the logits down to the range -3 to +3? If so, USCALE=. Did they want to show only the range -3 to +3 on some output plot or graph? Perhaps MRANGE= would help. Were they wanting to make the measures corresponding to extreme scores more central? Perhaps EXTREMESCORE= would do the job.
With questions like these, my answer touches on several different possibilities, expecting that, if I miss the target, another email will arrive from the questioner giving more context.
With all the recent travel, the Papers due to be reviewed or written have been piling up - my apologies to the authors and editors.
As far as I know, all urgent bugs have been fixed in the current versions of both Winsteps and Facets. But both have long wish-lists of enhancements. Microsoft is the big unknown. Their new 64-bit version of XP will no longer support DOS programs. That's fine - but I'm not sure what that means for batch mode operation of Winsteps and Facets. Microsoft are also discontinuing Visual Basic 6 in favor of Visual Basic .NET. But .NET does not seem to function properly under earlier Windows versions, and .NET may not support other functions currently implemented in Winsteps and Facets. Microsoft have done this to me before when they discontinued Professional QuickBasic and also when they discontinued Professional Fortran. Ultimately, both discontinuations turned out to be good for Winsteps and Facets, but they were traumatic at the time.
It is really great to have others picking up the workshop load. Ken & Nick tell me their Winsteps Workshop in Chicago a few days ago really hummed. Nick writes: You will be pleased to learn that some of our participants were having a "riot" with WINSTEPS graphing procedures. At a certain point, they left the main group to investigate DIF issues.
And, after that, Carol and Lidia conducted a Facets workshop. Carol writes: This was a really fun group to work with, and they seemed to develop their own group spirit over the two days (and are interested in staying in touch with one another).
Meanwhile, the Winsteps Summer Sale was humming along. It was an experiment motivated by Henry Ford. Before his time, automobiles were an expensive toy for the rich and technically oriented. He perceived the car to be the means of transport for the masses. For that to happen, he needed to produce a higher quality and much more reliable product at a much lower price. Hence the production line, more robust components, etc. And the Model T success story followed. Now we all have cars. Winsteps is already, in many respects, higher quality, more reliable software. It is regularly sold at a lower price. But what if the price was halved - down to the price of shareware? That was the Summer Sale. Initial reports from Digibuy, our software sales agent, are favorable - but what will sales look like for the rest of the year?
Sometimes the simplest questions are the most profound. The simple question prompts some thought: We need to restrict the range of our logits output to plus and minus 3. How can we do this?
That was all there was in the email. Did they want to rescale the logits down to the range -3 to +3? If so, USCALE=. Did they want to show only the range -3 to +3 on some output plot or graph? Perhaps MRANGE= would help. Were they wanting to make the measures corresponding to extreme scores more central? Perhaps EXTREMESCORE= would do the job.
With questions like these, my answer touches on several different possibilities, expecting that, if I miss the target, another email will arrive from the questioner giving more context.
With all the recent travel, the Papers due to be reviewed or written have been piling up - my apologies to the authors and editors.
As far as I know, all urgent bugs have been fixed in the current versions of both Winsteps and Facets. But both have long wish-lists of enhancements. Microsoft is the big unknown. Their new 64-bit version of XP will no longer support DOS programs. That's fine - but I'm not sure what that means for batch mode operation of Winsteps and Facets. Microsoft are also discontinuing Visual Basic 6 in favor of Visual Basic .NET. But .NET does not seem to function properly under earlier Windows versions, and .NET may not support other functions currently implemented in Winsteps and Facets. Microsoft have done this to me before when they discontinued Professional QuickBasic and also when they discontinued Professional Fortran. Ultimately, both discontinuations turned out to be good for Winsteps and Facets, but they were traumatic at the time.
