posted 10-29-2006 07:14 PM
In the continuing saga of attempting to understand how computer scoring programs work...I've continued to develop an Excell spreadsheet that can take hand scored data and test it for statistical significance using common algorithms.
Here is another .pdf link (17k) of a recent test, for which the young man failed and admitted to throwing a knife at another youth at school.
http://www.raymondnelson.us/qc/sample_sig_10-29-06.pdf
I interrogated on the minus one (-1) score before I put the numbers in to the spreadsheet.
This is a mixed issue test, at the request of the attorney who wanted to attempt to resolve concerns about verbal threats to stab the other youth, and where exactly the knife was thrown. (The knife was recovered, and its already clear to whom it belonged, and that it was unfolded.)
You can see the quantile plot are quite granular, making a difficult justification for parametrics.
Luckily, I've spent my most of my recent spare time building worksheets to calculate non-parametric equivalents of t-tests, hypothesis test, and anovas.
You can see the non-parametric scores are quite close to the parametric scores - without making assumptions about the distrubtion.
I still have more work to do on transformation and standardization of the data, from hand-scored values to machine scorable values.
I have yet to put the non-parametrics into the spreadsheet to score measurement values. Soon, I'll be able to investigate the concurrence between hand-scored and measurement scored data.
I've put the hand scored totals on the printout.
Here is another test, - a sex history test on a lifelong compulsive exposer. He's has multiple convictions for non-contact offeses, and was cut loose by a court decision in 2004 - regarding the legally of certain sentence lenghts. He's back now, with multiple charged incidents in multiple counties.
http://www.raymondnelson.us/qc/sample_sig_10-29-06_2.pdf
This test has an unresolved response to a question about physical sexual contact (sexual touching) with relatives or family members.
The Q-Q plots are good on this one - with nine relevant and nine comparison iterations. You can see the concurrence between hand-scores, parametric scores, and non-parametric scores.
I'm starting to like the nonparametric solutions better.
OK, enough brain damage for one weekend. I'm off to the gym for some treadmill and iron time.
r
------------------
"Gentlemen, you can't fight in here, this is the war room."
--(from Dr. Strangelove, 1964)