To anyone interested, that book can be downloaded for free with minimal effort.
Now, there are various issues that need to be taken into consideration when judging the worth of a meta-analysis. These can be found pretty easily with Google, but we can discuss them here if you’d like.
Beyond that, I don’t personally like throwing around meta-analyses in discussions. Not because I feel they are inherently worthless, but because I’m generally distrustful of anything that requires the manipulation of complex and diverse sets of data with nutty statistical software in order to synthesize results. That’s just my personal bias; doesn’t necessarily mean anything beyond my general skepticism of most publications and subpar education in mathematics.
However, this analysis only looked at eight studies. Since I’ll be done with finals tomorrow, if we can them in their entirety, we can use them as skill-building exercises as you read through your book.[/quote]
I see what youre saying, but with respect, meta-analyses represent one of the most important tools in our arsenal for being able to understand the mechanistic links between various treatments and outcomes. For a nice example, look at the explanation of why the Cochrane Collaboration logo is what it is:
No-one is saying that a meta-analysis cant be without fault. But what they can do is take a whole host of studies dealing with the same question and use all the data contained within to provide much more robust insight than any single trial could do. This is because some trials have low sample sizes, or low effect sizes, or big effect sizes with huge uncertainty (e.g. confidence intervals), and a meta-analysis can integrate over all this uncertainty to give clearer answers, while at the same time reducing the bias that any one study can contribute because there was something “special” about it (intentional or not). If we were to look at that special study in isolation, our conclusions might be totally different to the generalized outcome from the corresponding meta-analysis.