Quote:
|
Originally Posted by abaya
Well, I'm still waiting to hear what Jenny says is the standard p value in the field she's quoting. In anthropology, below 0.05 might be fine, but it varies for each field. That's the problem with statistics... you can prove that almost anything "holds water," if you want to.
|
My understanding is that medicine has stricter standards than the social sciences. Even then, p=0.002 is a fairly strict standard. That said, p< whatever doesn't automatically make something "hold water". There are a variety of other factors to consider when evaluating a study independent of the field: effect size, experimental design, composition of the sample, etc.
Also, I disagree that you can prove that almost anything "holds water" with statistics (and that it's a problem with statistics):
1) You can't prove anything anywhere in science with or without statistics.
2)The strength of statistics (and peer-reviewed journal articles) is that everything is there on the page. You can read how statistics were used to evaluate a given study. You can come to the conclusion that they used the wrong analysis or that some other aspect of the study was lacking.
3) The explicit nature of the study and the accompanying description of the statistical analysis used is a recipe of sorts for others to use to replicate the study. Perhaps the authors hypotheses were supported in this study, but will not be supported in subsequent replication attempts.
In regard to the study being discussed, there are a variety of other factors to consider other than the p value:
1) How big was the effect size? From the study, it appears that barbituate usage during the birthing process increases the likelihood of later barbituate addiction 1-7x. But what is the base likelihood? A 1-7x increase in likelihood might not be that much.
2) Has the study been replicated?
3) Have similar studies been done using samples taken from populations other than Swedes?
4) Are opiate addicts and their siblings a representative sample?
5) Are there other factors that account for greater amounts of the variance in barbituate usage in adulthood?
6) Have other studies shown contradictory effects? At least one of the links provided by Jenny led to a study by the same authors suggesting that barbituate usage during delivery leads to a lower risk of suicide later in life.
7) What are the consequences of not administering drugs when medically necessary? Are those consequences worse than an increase in likelihood of barbituate use later in life?
8) What percentage of births actually involve barbituate administration? And to the degree necessary to produce the effect described in the study? In order to achieve the 7x increase in likelihood, three doses of barbituates must be administered.
9) Barbituate during delivery is only a very small part of the larger issue of free-birthing. I have clearly become sidetracked in this thread by this issue.
My last comment on the study: The authors conclude that obstetric pain relief methods are preferable that do not permit substantial passage of drugs through the placenta. I can agree with that.
Cynthetiq, thanks for the infant mortality data. It was informative.