Bookmark me or the Baron will pull my heart plug thingy.

Wednesday, March 02, 2005

Spectrum Tests Part 3

Part 1 questioned such tests’ real world application.

Part 2 offered a tongue in check alternative test.

Part 3 offers a methodology which could be used to create better tests.

If you accepted the problems premised in parts 1 and 2 the solution to those problems will be no surprise.

Spectrum Test Creation Protocol:

Step 1. Find 100 of each of these groups:


a) conservatives

b) independents

c) liberals

d) whatever you like

Step 2

Have each of them write down the top 10 or 20 reasons that they are what they are.

Step 3

Compile, analyze and weight the results

Step 4

Create a test using the results

Step 5

Repeat step 1.

Step 6

Test the test.

Step 7

Adjust the test, correcting it according to how well the self-described of each group fit into each test category.

Step 8

Repeat step 1

Step 9

Repeat steps 6,7, and 8 until the test is accurate.

This protocol results in a self-correcting test. Once a template is created it would be VERY interesting to see how events CHANGED the results! It may seem impractical, given the hundreds or thousands of people necessary, but perhaps groups of 10 rather than 100 would suffice.

Maybe there already are tests out there created in such a manner. Actually the idea is so obvious I’m sure somebody is already doing it, for private if not public purposes. In fact I suppose political parties already use similar procedures to create their own polls. But, if so, why aren’t any such tests available on the net?


Post a Comment

<< Home