posted 11-04-2004 09:42 PM            

I am a new examiner interested in creating a database (probably using Microsoft Access) to track the tests that I conduct. I would like to use this database to quickly recall key information about past tests, and to track my own performance using empirical data rather than by guessing.

I was wondering if any of you "more seasoned" examiners can give me suggestions about what data points I should capture. I don't want to reinvent the wheel if somebody has a data system that works well for them.

So far I had the following data points in mind:

1. Case number
2. Name of examinee
3. Gender of examinee
4. Age of examinee
5. Type of issue (sex, burglary, etc.)
6. Opinion rendered
7. Date of exam
8. Test format used
9. Name of reqesting client, agency etc.

The question I would pose to veteran examiners is: "What is the one thing you wish you would have statistically tracked when you first began your polygraph career?" Any suggestions are appreciated!

--James

posted 11-05-2004 05:32 AM            

What software are you using? If you have Lafayette (maybe the others too, but I don't know), it will export the info from your PF file to Microsoft Access, and from there you can produce a numer of reports depending on what data you want to play with. It will take some time to fumble your way through the set up process, but once you get what you want you'll have a nice database.

posted 11-05-2004 07:46 AM            

I'm currently converting to MySQL as Access does not do what I want.

Your data fields look OK.

Somewhat controversial but you might also track ethnicity - generalizability of interpreted data depends on the representativeness of the sample.

As I work mostly in PSCOT, where many folks have mental and medical health issues, I also track medications.

I also track the last polygraph date and the number of prior polygraphs (ranges from none to 50).

I also suggest you create fields for mixed-issue and single-issue tests - and code the number of questions. These are empirical parameters that affect the test outcome to some degree.

You might also code for diagnostic (specific issue test) in response to known incidents, and tests that are conducted in the absence of a known incident (see Krapohl and Stern, 2003)

In my database is a list of referring agencies and every client I'v ever tested. I keep a description of the case background and family demographics, and for PSCOT cases a description of the instant offense.

In a few moments I can tell you the number of tests I've run - specific, PE, sex history, maintenance. Proportion of deceptive, non-deceptive, inconclusive, and PNC results. I can also sort this by referring agent or juristiction (some are doing a much better job than others.)

One thing to keep in mind is that where you work will affect your results (selection bias). When I work in prison settings, I see more victims, more medications, much more messy data, and sometimes poorer motivation. In more afluent and better educated neighborhoods - more countermeasures. Treatments programs that serve employed married folks with kids - pretty easy to motivate for truth. Secured treatment programs for disturbed juveniles - elevated proportion of sexual compulsives, chronic masturbation, and reckless (emerging) personality disorders (this makes sense as criminal justice data tells us consistently that early age of onset indicates greater risk levels and greater lifetime problems - this is antithetical to the social-workers assumption that early intervention improves outcome as those youths requiring early intervention are the same ones demonstrating that early onset).

Geeze if you didn't know better you'd think someone were doing a study and writing a dissrtation...

I can sort test results by the number of test questions, and there is an interesting relationship between the number of (mixed) test questions and the proportion of inconclusive or unresolved tests - this might be estimated mathematically by taking the inverse of the inversed (yes) inconclusive rate (5-10%) raised to the exponent of the number of test questions. Or, INC-All = (1 / INC) EXP NUMB-RELEVANT-QUESTIONS ---- sucks without a math editor, but try it.

Inconclusive (unresolved tests) rates can be predicted to go up with the number of mixed issues (questions).

I just queried my data for PCSOT tests and my overall inconclusive rate is under 5% for sex history, maintenance and instant offense. I didn't query the confirmation rate, but I know anecdotablly most DI cases are resolved.

I can export all this to a report in minutes.

Lots of fun...

Hey Ralph, how about a little message icon with a spinning propeller-hat????