Thread: Identifying a convex bevel caused by extensive pasted stropping

Originally Posted by edhewitt

Jimbo, do you mean that it is like trying to describe the flavour of an orange to someone who has never eaten an orange?

I suppose it's a bit like that Ed, but not entirely.

Basically, without getting too technical, what is going on is what's known as an ordinal scale. It is not the same as a numerical or ratio scale, where we can say that 4 is twice as big as 2 and so on, but it is not quite as arbitrary as a completely categorical system (where we can interchange the labels so long as we remember what they mean - eg we might label patients in a drug trial as A = drug group and B = control group, but the labels can be swapped - or are arbitrary- so long as we keep track of which label refers to which group).

So the HHT-x is ordinal - the x's are categories, but there is order associated with them ie 1 < 2 ... < 5. Now there's nothing wrong with that at all from an internal consistency perspective, obviously. I know what I mean by the numbers, and in fact as I gain more experience using the scale I will most likely jigger and readjust things until I get the scale just how I like it and it will be quite useful for me.

Problems arise, however, with inter-person use. There are two main issues: location and scale. I think I can demonstrate both of these diagrammatically, but don't hold me to that

The location issue is where two people start the scale. My HHT 1, because of my hair type, may in fact correspond to your HHT 2 for example. That is, on the same edge my hair would, say, simply split whereas your hair may in fact cut through. So we'd be out of sync like this, for example:

1 *** 2 *** 3 *** 4 *** 5
.........1 *** 2 *** 3 *** 4 *** 5

The other problem is scale. Ordinal scale cut offs cannot be assumed to be equi-distant (as I pictured above). Different individuals may have different "elasticities" between cut offs:

1 ***** 2 ** 3 * 4 ****** 5
1 ** 2 **** 3 ** 4 **** 5

And of course there could be combinations of scale and location differences between people too.

There's a rather large literature in the mathematical and, more particularly statistical, literature on this kind of thing. As a general rule, any proposed measure of something using this kind of scale needs to be assessed both for its intra-person reliability (which for HHT-x I think is probably quite good after a bit of calibration) and its inter-person reliability (which is where I think the HHT-x really falls down).

Of course, poor inter-rater reliability of the HHT-x is pure (educated) conjecture on my part. I could quite possibly be very wrong. The only way to find out would be to conduct inter-rater testing which would involve many raters assessing the HHT-x behaviour of an identical set of edges independently of each other. If anyone wants to do that I'd be more than happy to advise the experimental design, and I'll even throw in the analysis and waive my usual consultation fee

Please also note that unless coticules somehow impart hair homogeneity to everyone who hones with them, the problems I've outlined above are not non-coticule specific I'm afraid.

James.