Posted by SLS on January 6, 2008, at 17:00:02
In reply to Re: correlation does not imply causality » Jamal Spelling, posted by seldomseen on January 6, 2008, at 16:08:03
> "For example, a similarly comprised untreated group might have a remission rate of 40% after 56 weeks, which would imply that the STAR*D algorithm really only works for 26% of patients. Then the abstract would have to read something like "The study demonstrates that 26% of depression patients can be brought into remission by 56 weeks as a result of following the STAR*D algorithm". And that doesn't sound as good."
>
> Actually that's not how that data should be interpreted. If 40% of patients spontaneously remitted and 67% remitted on with drugs, then 67% still remitted on drugs.
>
> The only valid conclusion is that more subjects remitted on drugs than with no treatment.
>
> We don't know how many people the drugs actually helped, just that more remitted while on them.This is where calculations of statistical power come in to play, no?
> I think everyone(and I'm not on one side or the other) have to be very careful how they interpret the data.
I may be wrong, but it seems to me that we do not have many PB members whom are well-versed in statistics. Of course, people who have an emotional investment in winning an argument such that they must defend an agenda rather than search for a truth. In my opinion, some people dissemble in order to win arguments. It becomes a well-honed skill.My point is that most of us lack the expertise to design, implement, and offer interpretations at the level of professional investigators. That's why I feel we are blessed to have members with various expertises to enhance our ability to locate, review, interpret, critique, and offer conclusions for an observed phenomena.
Thank you all.
- Scott
poster:SLS
thread:804126
URL: http://www.dr-bob.org/babble/20080105/msgs/804671.html