Tuesday, February 07, 2006

Only a sampling bias?

Scientific research is full of traps, especially regarding the use of biased samples. For example, the first polls in the US started in the early 1900, I think. And they got the prediction completely wrong, because they took voters out of the phone directory. But in early 1900, only wealthy people had phones and so they completely missed the non-phone population's political views which not surprisingly happen to be the opposite. And the candidate low in the polls won with a landslide. This post is about a possible sampling bias that creates an effect that looks intriguing and surprising, but is probably not.

Have a look at this paragraph:

"In the group of stuttering preschool children in the study summarized above (Yairi&Ambrose, 1999) their expressive language abilities were measured. The group with early onset stuttering, who entered the study at age 2-3, showed syntactic abilities and length of utterances well above what was expected for... In fact, in some aspects the language abilities in this group were on a level with the norms for 2 years older children. This was true both for children who recovered and for children who persisted to stutter. Children who entered the study at a later age showed language abilities at about age expectations, except for the group of children with persistent stuttering in the oldest age group (entering the study at age 4?5), whose abilities were somewhat below the norm. These results indicate that children with early onset stuttering tend to show precocious learning of language."

So does early language development causes stuttering? This is also what I was asking myself, but it didnt make too much sense to me. I heard many people quoting this study and another one by Rommel et al. (In Montreal, I went to see a Radiohead concert with Rommel.) But now I believe that this finding might be nothing else than a sampling bias effect. On the STUTT-L mailing list, Don Luke made an interesting observation: "This would make sense. All these children began stuttering upon reaching the same point in their development of language skills."

Here is my argument:

1) Let us assume that stuttering starts at a certain stage in the development (of language and speech skills). This is a very reasonable assumption to make.

2) Children naturally fluctuate in their development; some are earlier than others, but starting early does not mean having superior abilities later.

3) So the early-starters hit the critical stage first where they are vulnerable to stuttering!

4) So early-onset stuttering kids have early development and therefore superior-than-peers language skills at that age (but not necessarily in adulthood)

So I hope you can see that this is a pure statistical effect that has nothing to do with a direct casual link to stuttering. There is only a correlation. It might also be that late-starters go through that stage too late to start stuttering for whatever reason.

DISCLAIMER: I have actually not read the articles in detail, so I might start from a wrong premise, but the stats argument must be correct as it is maths.

No comments: