While there are some exceptions, in almost every situation putting many studies with divergent treatment methods into a single meta-analytic model gives you an entirely meaningless result
-
-
Replying to @GidMK @chad_senger
Think of it like an average (because it essentially is one) - if one study measured weight, another height, and yet another blood pressure, averaging the results from them all would give you a number with narrow confidence intervals that had literally no meaning
1 reply 0 retweets 1 like -
Replying to @GidMK
Your straw-manning again. Yes, if you measure completely different things it’s worthless. But if the studies are similar, you can eliminate noise and outliers.
1 reply 0 retweets 2 likes -
Replying to @chad_senger
If you have multiple very similar small trials that measure the same thing in a very robust way, meta-analysis is great at giving you a more certain estimate. But if the studies are bad, or very divergent, it gives you no more certainty than you have to begin with
2 replies 0 retweets 1 like -
Replying to @GidMK
This is of course true. But not unique to meta analysis, so it’s not clear why your focusing on them. A poorly run large scale double blind randomized control trial is just as worthless if it’s run poorly. A meta analysis could at least overcome a bad study with the aggregate
1 reply 0 retweets 0 likes -
Replying to @chad_senger
That is not true actually. Meta-analyses can be biased by bad studies in a number of ways, often to do with the weighting. Sometimes a single bad study can push the entire model in a single direction
1 reply 0 retweets 0 likes -
Replying to @GidMK @chad_senger
The rest of your statement confuses me. I talk all the time about bad study design, but *this* conversation is about meta-analysis
1 reply 0 retweets 0 likes -
Replying to @GidMK
No it’s not, it’s about poorly run studies and bad data. You said it yourself, meta analysis’ are very valuable with good data properly run studies. The flaw you are picking on is not in the meta analysis at all, it’s in bad studies.
1 reply 0 retweets 0 likes -
Replying to @chad_senger
The point I was making is that meta-analysis does not "correct" for bad data in any way, nor does it resolve underlying issues with the research
1 reply 0 retweets 0 likes -
Replying to @GidMK
And no one is saying it does. But it only takes a cursory understanding of meta analysis to know that if you pool studies with different approaches/treatments, in different populations/ cultures... you are going to dampen the effect those outside influences have on your result
1 reply 0 retweets 0 likes
That is simply incorrect. Entirely wrong
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.