DEAR PROFESSOR STRAIN... Confounded by factors?

From: Bodil Carlsson [mailto:eiriasverchjanis@hotmail.com]
Sent: Tuesday, November 01, 2011 3:37 PM
To: George M Strain
Subject: Confounded
 
Dear Prof. Strain,

I´m sorry to plague you with ignorant questions once again, but once again I´m baffled. I have just read the full report of the 2006 Border Collie study by Platt, Freeman et al, Prevalence of Unilateral and Bilateral Deafness etc. I fully appreciate what you said about the dogs included not being genotyped and share now your view that this is anything but a minor problem - if no genotype, how do we know what we are talking about? How do we know what it is we see?

What disturbs me is this. In the concluding paragraph on Prevalence Estimates, authors say: "the generalized logistic regression results emphasize the high risk of deafness for BC:s with a merle coat (prevalence of of 7.9%, upper 95% confidence limit of 13.3%), blue eyes (prevalence of 9.3% for 1 blue eye and 23.9% for 2 blue eyes; corresponding upper 95% confidence limits of 15.7 and 40.0 respectively) or excess white on the head (prevalence of 36,3%, upper confidence limit of 55.3%.)"

I´m not familiar with the intricacies of logistic regression models. But I read this as meaning that  deafness in phenotypic merle was present in 7.9 with an upper probability limit of 13.3% ... whereas deafness in the genotypically undefined phenotypic sign of "excess white", meaning more than 50% of head white, was  present very much more often - ???
I e, there are two other phenotypic traits which carry a higher risk than phenotypic merle, and although each one of them MAY be an expression of the merle gene, each one may also be the expression of an S allele, as authors also claim that "all Border Collies are homozygous for the sw and sp alleles" and then conclude  "that the S locus is not thought to be involved in the regulation of deafness in Border Collies." How can one conclude that?

I would very much have liked to see, first, by genotype how M and m corresponded with deafness and at least as much how genotyped  S alleles corresponded. Am I totally off the track in thinking this?

Once again, very grateful if you have the time and patience to answer.

Regards,
Bodil Carlsson


confounding factors: uttryck för en form av begreppsförvirring i en studie, där saker som man tror är oberoende av varandra i själva verket hänger ihop, eller där ett speciellt urval av försökspersoner ell likn ger resultat som inte gäller om man hade valt att titta på ett större eller annorlunda urval. Här lite av en ordlek, för antingen är studien eller jag, som frågar, confounded av faktorerna jag frågar om.