Lol, fine: on a scale of 1 to random, this is pretty low down. I agree that they've done a good job reporting, but I personally don't think you can call the results of this procedure "random" in the common technical sense of the term
-
-
Replying to @GidMK
What's your scale and population? If we're saying 10 is true randomization, and we are including all studies and designs (including completely obs/non-randomized, I am still putting this at like a 9 on the randomization scale.
1 reply 0 retweets 2 likes -
Replying to @NoahHaber
Nah I'm only looking at RCTs here where people have attempted some kind of randomization. This is one of the worst methods I've ever seen
1 reply 0 retweets 0 likes -
Replying to @GidMK
Ok, relative to other randomization schemes, the procedure is pretty bad. Relative to other designs (including non randomized), its pretty good. I feel like maybe we should focus more on those other really bad (and far more common) designs.
1 reply 0 retweets 2 likes -
Replying to @NoahHaber
Dunno, most non-random designs are pretty open about the issues with causal inference. Bad randomization is used to infer direct causality despite the issues with bias
1 reply 0 retweets 0 likes -
Replying to @GidMK
Not sure I agree with your assertion about relative openness of bias in non-randomized studies. But again, what's the bias here? Interpretation changes due to non-blonding, sure, but I am not seeing any serious bias (in expectation) from this procedure.
1 reply 0 retweets 1 like -
Replying to @NoahHaber @GidMK
*non-blinding. Non-blonding is a totally different technical concept.
1 reply 0 retweets 1 like -
Replying to @NoahHaber
Sorry, I should've said potential bias. It's impossible to know what bias may have occurred, which I'd argue is also a weakness. At least in cohort studies we have a good idea of the major biases!
1 reply 0 retweets 0 likes -
Replying to @GidMK
We might be looking at very different cohort studies here, but I don't think I agree with that assertion either. There is also a difference between not having any reason to believe there is bias, and (not) knowing what biases may exist.
1 reply 0 retweets 0 likes -
Replying to @NoahHaber @GidMK
Put another way, I'd take this weirdly randomized n=43 person RCT (assuming no other major issues, which is admittedly a big assumption) over an otherwise similar n=43 cohort study basically any day.
1 reply 0 retweets 0 likes
I'd say it depends entirely on the cohort study, but either way I'd rather have a properly randomized trial given how easy (and free) that is to do!
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.