This page describes a way of testing the validity of information transfer through idiographic assessments. The key feature of the method is that if a judge can match four or more idiographic descriptors to the people or objects they describe, then the chances of this having arisen by chance alone are p<.05.

A paper showing the maths of this is:

Evans, C., Hughes, J. & Houston, J. (2002) *Significance testing the validity of ideographic methods: a little derangement goes a long way*. British Journal of Mathematical and Statistical Psychology, **55**:385-390.

I've mounted a very quick introduction to the method in PDF. That's a variant of a powerpoint poster given at the Chicago International meeting of the Society for Psychotherapy Research in 2000.

In the unlikely event that you can't read that there's a horrid HTML version as spat out of M$ Powerpoint in very nasty html. Assuming you're a M$ user, then a more recent version of Powerpoint spits it out more usably as an mht file this works fine in M$ Internet Explorer but which doesn't work for me in Firefox and I don't know about other browers.

Table 1 from that paper tabulates the probabilities of various scores for various *n* (Hm, horrid HTML, I must find a better way to get from Word to HTML than Word.)
If you observe a score of, say 6 correct matches across 8 sets of data,
you can read that the probability of hitting exactly six by chance alone
was .0007, and that the crucial probability of hitting six or more, is
also .0007 to any sensible accuracy (since the probability of getting
eight of eight by chance alone is a .00002 which adds little to .0007).
Incidentally, there is no possibility of scoring 7 here since it's
always impossible to match *n*-1 from *n*: if you've got the
first *n*-2 right, you either get the last two the right way round or the wrong way around.

In addition I've mounted a few tools that can give you these probabilities for any arbitrary n:

- interactive table generator which gives tables like that above for any
*n* - interactive calculator which gives p for any
*r*matched from any*n* - The programs behind those two interactive calculators are based on a program in R (tested in R 1.7.1 under Windoze2k and 1.8.0 on Linux) to give exact p for any score from any (sensible) n
- output from that R program (scroll to the end to see the table from all.derangements(8) and p.derange.score(6,8))
- S+ program (tested in S2000 under Windoze2k) to give exact p for any score from any (sensible) n
- SAS/IML program to do likewise will follow when I get it back off tape (don't hold your breath, don't know when I'll have time to fix this!)
- C++ program ditto

Chris Evans,

Rampton Hospital, Retford, Notts. DN22 0PD Britain

by telephone at [+44|0] 1777 247242

and 'fax at: [+44|0] 1777 247213