méta-regression par un modèle polynomial d'ordre 2

14:54 Mar 13, 2013 

I would like to add that it would be best if the exact nature of the "meta" in meta-regression were defined somewhere along the way. I.e., I'm not strictly convinced that the meta-regression is exactly as defined in Wikipedia. If the precise nature of the regression is otherwise desribed/defined, and this is the only instance of "meta-regression" I'd be half inclined to drop it in the translation.

09:09 Mar 12, 2013 

I don't quite see the point you're making. Regression (and meta-regression) models are common enough, essentially a least squares fit to the data. A parabola (order 2 polynomial) will fit most lightly curved data over a sufficiently restricted range, but the function to be fitted (the model) could be almost anything - trig, log, whatever; it's just that polynomials are simplest to work with.
And yes, accuracy is important, but so also is what you disparagingly refer to as its "aesthetic".

22:57 Mar 11, 2013 

I've seen few times the meaning of a ST slightly or not so slightly twisted or even "finir dans le décor" just in order to satisfy some esthetic/style imperative about "avoiding repetitions". What is more important: getting the meaning right or "how it's going to sound"?
The mathematical model in the ST is clearly labelled "un modèle polynomial d’ordre 2" (that's how data is processed) and there's no point amalgamating that with "méta-regression" that's got to do with how data is collected - through a meta-study.

22:30 Mar 11, 2013 

Yes, definitely too many "models". A more natural way to express it might be simply to say "a quadratic meta-regression model for the prevalence....". Nothing wrong with "order 2", but "linear", "quadratic" and "cubic" are often used in preference to the more cumbersome "order n (1, 2 or 3)", unless n is greater than 3.

18:34 Mar 11, 2013 

Also, your question is a bit confusing because the text doesn't actually say "méta-regression par un modèle polynomial d'ordre 2"

18:31 Mar 11, 2013 

Seems fine to me, although I think 2nd/second order is better than "of order 2"

18:25 Mar 11, 2013 

I'm not a statistician, but surely a literal translation is fine? A meta-regression showing the prevalence of ... using a polynomial model of order 2? You have to find a way of avoiding the repetition of "model", which I think is a mistake in the French.