werner@aecom.UUCP (Craig Werner) (05/25/85)
Before any more of this nonsense gets posted:
First seifert@mako didn't get the statistics. Now ems@amdahl is
blaming everything on miniscule amounts of dye.
OK, this is getting ridiculous. If the packaging caused cancer, then
you would expect people in the experimental group to get cancer, too. Also,
how much difference will one pill make in total chemical intake (it was one
pill/day) Two, chemical carcinogens tend to be organ specific - in this study
5 patients got five different cancers -- all of them common in the population.
We CAN be sure. They didn't get the cancer by increasing their total dye
intake by .0001% of daily intake.
> What do you expect? Watermelons are out of season!
This will change. But the question is, did anyone figure out its source?
DO NOT, I PLEAD, DO NOT, respond to this article with a POSTING!!!
--
Craig Werner
!philabs!aecom!werner
"The world is just a straight man for you sometimes"eric@osiris.UUCP (Eric Bergan) (05/26/85)
> First seifert@mako didn't get the statistics. Now ems@amdahl is > blaming everything on miniscule amounts of dye. > > OK, this is getting ridiculous. If the packaging caused cancer, then > you would expect people in the experimental group to get cancer, too. Also, > how much difference will one pill make in total chemical intake (it was one > pill/day) Two, chemical carcinogens tend to be organ specific - in this study > 5 patients got five different cancers -- all of them common in the population. > We CAN be sure. They didn't get the cancer by increasing their total dye > intake by .0001% of daily intake. > Actually, after reading some of these postings, I decided to ask my boss (ex-head of Oncology at Johns Hopkins Hospital). He said that at various times, there have been movements within the medical community that did believe that cancer was psychosomatic. Therefore, if the patients for some reason believed that the placebo could cause cancer, it might cause them to have cancer. He does not personally fall into this camp, but he did not laugh at it, either. -- eric ...!seismo!umcp-cs!aplvax!osiris!eric
ems@amdahl.UUCP (ems) (05/31/85)
> Before any more of this nonsense gets posted: Excuse me, but I was not notified when you were appointed censor-of-net-med. Might I see the official document that empowers you to pass judgement on others? > First seifert@mako didn't get the statistics. Now ems@amdahl is > blaming everything on miniscule amounts of dye. First, as I remember it, the question was asked (paraphrase) - or is the sample size too small ... - This would indicate to me an understanding of the basic statistics, but a disire to avoid digging out the stats book to calculate the sample size needed for a small standard deviation when other important factors were not available to the person asking. It did not show, to me, that mako "didn't get" the statistics, but rather that they understood the statistics and were asking for an estimate of standard deviation for that sample size. Second, I don't remember 'blaming everything' on dye. I do remember *ASKING* if you could state with certainty that the dye was not implicated, along with other potentially overlooked items. I do believe that *MOST LIKELY* the effect was a statistical anomaly; however, I also appreciate that we cannot really *KNOW* that it was. That was my point. We can make reasoned guesses based on the best double blind studies using the tools of statistics, but we cannot truely *KNOW*. Using Occam's Razor, we will choose one *THEORY* to beleive in as true, until it is superceeded by another. When an anomaly such as this happens, the proper course is to look for an independent variable that was thought to be a constant. Rember Legionairs Disease? Statistic or an unknown effect? It may be the dye, the office that the placebo was given in (If different from where the test drug was given - there are cases of radiation leaking through walls ...), a lack of 100% purity in the starch preparation, etc. If none is found it does not prove randomness, it proves either randomness *OR* that we have not found the cause. We then assume for convenience that it was randomness, pending further data. As for the amount of dye being small: What dose level of dioxin is needed to induce cell changes? Carcinogens are often noted for the very low levels needed to cause problems. What organic chemical reaction produces 100% (integer, not rounded float) pure product? (Please consult your organic chem texts, stats books, and physics texts - isotopes you know - before answering) My point remains: you can not know. You can assume. > OK, this is getting ridiculous. Yes, it was. > If the packaging caused cancer, then > you would expect people in the experimental group to get cancer, too., Only if the packaging were identical. You have not stated that this is so. While it is inferred, it is not a given. > Also, > how much difference will one pill make in total chemical intake (it was one > pill/day) That depends on what is in the pill. All chemicals are not created equal. > Two, chemical carcinogens tend to be organ specific - in this study > 5 patients got five different cancers -- all of them common in the population. A good point. This was missing from your earlier posting and is valuable information. But please notice the phrase 'tend to be'. We are agreeing here that they need not be. You are assuming that there is not a chemical carcinogen because of a trend, not a 100% rule. This still leaves room for uncertainty. > We CAN be sure. They didn't get the cancer by increasing their total dye > intake by .0001% of daily intake. Depends on the dye and any contamination. Were the placebos prepared in the same place and at the same time as the drugs? Were *ALL* of the factors held constant? Was there *NO* chance for contamination at *ANY* step? It has often been true that the perplexing 'statistical' anomalies overlooked by one researcher were the seed of brilliant discoveries when investigated by another. > DO NOT, I PLEAD, DO NOT, respond to this article with a POSTING!!! Why not? You have raised some issues that need responding too. Is there a reason not to do it in the same public forum where they were raised? Please forgive my usual compliment of spelling, diction, style, grammar, etc. errors. -- E. Michael Smith ...!{hplabs,ihnp4,amd,nsc}!amdahl!ems Tilapia Zilli is the way and the light. This is the obligatory disclaimer of everything.
ems@amdahl.UUCP (ems) (06/04/85)
In reading my prior posting I noticed that there was a possible interpretation of the statistical discussion that was not what I had intended. That is, I refer to the size of the standard deviation in a way that might be taken as meaning that is could be derived from the sample size. This was not the interpretation I had intended. The standard deviation of a population is fixed and independent of the sample size. What I should have said is that it was desired to find the confidence interval (or error factor) for this observation given a small sample size and a presumed small standard deviation. I hope that this brief exposition on statistics has not caused too many of you to fall asleep at your terminals, but I felt compelled to post the observation myself before 20 to 30 math majors posted their stats texts... :-) The offending lines follow: > > First, as I remember it, the question was asked (paraphrase) - or is > the sample size too small ... - > This would indicate to me an understanding of the basic statistics, > but a disire to avoid digging out the stats book to calculate the > sample size needed for a small standard deviation when other > important factors were not available to the person asking. It did > not show, to me, that mako "didn't get" the statistics, but rather > that they understood the statistics and were asking for an > estimate of standard deviation for that sample size. > > -- E. Michael Smith ...!{hplabs,ihnp4,amd,nsc}!amdahl!ems Tilapia Zilli is the way and the light. This is the obligatory disclaimer of everything.
edhall@randvax.UUCP (Ed Hall) (06/04/85)
Methinks he doth protest too much. Science would never progress if it expended all its energy trying to rule out one-in-a-million chances. Granted, a few things might get missed over the years, but if we spent our time straining at gnats, no progress at all would occur--which in the long run may have a far worse effect than prematurely accepting a discovery as fact. At least in the latter case the mistake will probably come to light eventually. It's the old economic law of diminishing returns, and isn't peculiar to science or medicine. There is no such thing as a ``perfect'' study, and no such thing as statistical certainty. The idea is to reach reasonable certainty, and then move on, trying to build on what came before. We don't have forever, and at least in medicine, lives depend on timely solutions. -Ed Hall decvax!randvax!edhall
ems@amdahl.UUCP (ems) (06/07/85)
> Methinks he doth protest too much. > Perhaps, but this *IS* the net ... :-) > Science would never progress if it expended all its energy trying to > rule out one-in-a-million chances. Granted, a few things might get > missed over the years, but if we spent our time straining at gnats, no > progress at all would occur--which in the long run may have a far > worse effect than prematurely accepting a discovery as fact. At least > in the latter case the mistake will probably come to light eventually. > Very good points. I would add, though, that the 'probably come to light' depends to some extent on a willingness to re-examen old ideas and to overturn accepted dogma. These traits have not come easily to the academic professions, medicine in particular. > It's the old economic law of diminishing returns, and isn't peculiar > to science or medicine. There is no such thing as a ``perfect'' > study, and no such thing as statistical certainty. The idea is to > reach reasonable certainty, and then move on, trying to build on what > came before. SO TRUE! But we need to remember that we are only dealing with 'reasonable certainty' and not with *FACT*. We need to keep our eyes and minds open for the anomalies that give the lie to our dogma. > We don't have forever, and at least in medicine, lives > depend on timely solutions. > An excellent point. While using that solution, though, we need to keep asking our selves: What are we assuming that may be wrong? Remember how many years science was set back by folks who would not believe what Galilaeo (sp?) could see. There are many more examples. -- E. Michael Smith ...!{hplabs,ihnp4,amd,nsc}!amdahl!ems This is the obligatory disclaimer of everything. (Including but not limited to: typos, spelling, diction, logic, and nuclear war)