 |  |
 |  |  | |||||||
|
--- Prominent cardiologist criticizes drug ads
"Herman Rubin" news:cr736m$jcs@odds.stat.purdue.edu... > In article > reeder > >On 30 Dec 2004 16:00:24 -0500, hrubin@odds.stat.purdue.edu (Herman > >Rubin) wrote: > > >>Do you have any idea what statistical significance means? > > >>I doubt that Topol does. > > >I'm sure Dr. Topol must have taken Statistics 101 in college: > > Statistics is not a normal part of the pre-med program. > > Also, methods courses make it difficult to understand > concepts, and Statistics 101 is almost always a methods > course. Any attempt to teach it otherwise would cause > students who do not have a fairly strong mathematics > backgroumd to get the material in another department, > often under the name "research methods". > > I have heard of someone submitting results to a medical > journal being told to collect more data until the > results were significant at the 5% level and then the > paper would be published. This invalidates much of > the religious use of statistics in medicine. Which comment invalidates the use of the .05 level Herman?
In article <6ZFBd.14627$RH4.7093@newsread1.news.pas.earthlink.net>, George Conklin >"Herman Rubin" >news:cr736m$jcs@odds.stat.purdue.edu... >> In article >> reeder >> >On 30 Dec 2004 16:00:24 -0500, hrubin@odds.stat.purdue.edu (Herman >> >Rubin) wrote: >> >>Do you have any idea what statistical significance means? >> >>I doubt that Topol does. >> >I'm sure Dr. Topol must have taken Statistics 101 in college: >> Statistics is not a normal part of the pre-med program. >> Also, methods courses make it difficult to understand >> concepts, and Statistics 101 is almost always a methods >> course. Any attempt to teach it otherwise would cause >> students who do not have a fairly strong mathematics >> backgroumd to get the material in another department, >> often under the name "research methods". >> I have heard of someone submitting results to a medical >> journal being told to collect more data until the >> results were significant at the 5% level and then the >> paper would be published. This invalidates much of >> the religious use of statistics in medicine. > Which comment invalidates the use of the .05 level Herman? The use of the .05 level, or the p-value which is a little more detailed, tells me nothing about the effectiveness of the new treatment. I do not need any information to tell me that there is a difference, so why should I consider a test that there is no difference to be of any import? Also, there is a lot of use of meta-analysis recently in medicine. For this to be as stated, it must include all previous studies, not just those published. While some studies in which some "non-significant" effects for a particular criterion get published because something else is significant, it will still be the case that the sample can be highly biased. Large numbers are no protection. -- This address is for information only. I do not claim that these views are those of the Statistics Department or of Purdue University. Herman Rubin, Department of Statistics, Purdue University hrubin@stat.purdue.edu Phone: (765)494-6054 FAX: (765)494-0558
In article hrubin@odds.stat.purdue.edu (Herman Rubin) wrote: > Also, there is a lot of use of meta-analysis recently in > medicine. For this to be as stated, it must include all > previous studies, not just those published. While some > studies in which some "non-significant" effects for a > particular criterion get published because something else > is significant, it will still be the case that the sample > can be highly biased. Large numbers are no protection. If this were to determine a gold or any other mine it would be funny. Can you imagine this bore hole says 150 meters of a rich deposit. Just forget the other borehole data. Thanks Herman! Bill who wonders what the distribution of common sense is these days. -- Zone 5 S Jersey USA Shade Serious Vision Problems like Starghartıs ? --> http://www.ocutech.com/
"Herman Rubin" news:crbqor$4tmg@odds.stat.purdue.edu... > In article <6ZFBd.14627$RH4.7093@newsread1.news.pas.earthlink.net>, > George Conklin > > >"Herman Rubin" > >news:cr736m$jcs@odds.stat.purdue.edu... > >> In article > >> reeder > >> >On 30 Dec 2004 16:00:24 -0500, hrubin@odds.stat.purdue.edu (Herman > >> >Rubin) wrote: > > >> >>Do you have any idea what statistical significance means? > > >> >>I doubt that Topol does. > > >> >I'm sure Dr. Topol must have taken Statistics 101 in college: > > >> Statistics is not a normal part of the pre-med program. > > >> Also, methods courses make it difficult to understand > >> concepts, and Statistics 101 is almost always a methods > >> course. Any attempt to teach it otherwise would cause > >> students who do not have a fairly strong mathematics > >> backgroumd to get the material in another department, > >> often under the name "research methods". > > >> I have heard of someone submitting results to a medical > >> journal being told to collect more data until the > >> results were significant at the 5% level and then the > >> paper would be published. This invalidates much of > >> the religious use of statistics in medicine. > > > Which comment invalidates the use of the .05 level Herman? > > The use of the .05 level, or the p-value which is a little > more detailed, tells me nothing about the effectiveness of > the new treatment. I do not need any information to tell > me that there is a difference, so why should I consider a > test that there is no difference to be of any import? > The test tells you whether a drug is any better than nothing. Drug companies seldom test a drug to see if it is better than an existing one, or a cheaper one. > Also, there is a lot of use of meta-analysis recently in > medicine. For this to be as stated, it must include all > previous studies, not just those published. While some > studies in which some "non-significant" effects for a > particular criterion get published because something else > is significant, it will still be the case that the sample > can be highly biased. Large numbers are no protection. Well, yes, drug tests which show no significant result are not published. The same is true in social science research too. The result is that there is a bias and a whole lot of wasted research. No significance is just as important as significance, and you are right in pointing out a bias here. > -- > This address is for information only. I do not claim that these views > are those of the Statistics Department or of Purdue University. > Herman Rubin, Department of Statistics, Purdue University > hrubin@stat.purdue.edu Phone: (765)494-6054 FAX: (765)494-0558
In article George Conklin >"Herman Rubin" >news:crbqor$4tmg@odds.stat.purdue.edu... >> In article <6ZFBd.14627$RH4.7093@newsread1.news.pas.earthlink.net>, >> George Conklin >> >"Herman Rubin" >> >news:cr736m$jcs@odds.stat.purdue.edu... >> >> In article >> >> reeder >> >> >On 30 Dec 2004 16:00:24 -0500, hrubin@odds.stat.purdue.edu (Herman >> >> >Rubin) wrote: >> >> >>Do you have any idea what statistical significance means? >> >> >>I doubt that Topol does. >> >> >I'm sure Dr. Topol must have taken Statistics 101 in college: >> >> Statistics is not a normal part of the pre-med program. >> >> Also, methods courses make it difficult to understand >> >> concepts, and Statistics 101 is almost always a methods >> >> course. Any attempt to teach it otherwise would cause >> >> students who do not have a fairly strong mathematics >> >> backgroumd to get the material in another department, >> >> often under the name "research methods". >> >> I have heard of someone submitting results to a medical >> >> journal being told to collect more data until the >> >> results were significant at the 5% level and then the >> >> paper would be published. This invalidates much of >> >> the religious use of statistics in medicine. >> > Which comment invalidates the use of the .05 level Herman? >> The use of the .05 level, or the p-value which is a little >> more detailed, tells me nothing about the effectiveness of >> the new treatment. I do not need any information to tell >> me that there is a difference, so why should I consider a >> test that there is no difference to be of any import? > The test tells you whether a drug is any better than nothing. Drug >companies seldom test a drug to see if it is better than an existing one, or >a cheaper one. I have seen them with comparisons with other drugs. It definitely does NOT do this. Anyhow, that is not the proper question to ask; the question to ask is whether it is worth taking by some people, who might be able to make the decision given the information. It is not for the FDA to make such decisions, but rather to see that the information is provided. >> Also, there is a lot of use of meta-analysis recently in >> medicine. For this to be as stated, it must include all >> previous studies, not just those published. While some >> studies in which some "non-significant" effects for a >> particular criterion get published because something else >> is significant, it will still be the case that the sample >> can be highly biased. Large numbers are no protection. > Well, yes, drug tests which show no significant result are not published. >The same is true in social science research too. The result is that there >is a bias and a whole lot of wasted research. No significance is just as >important as significance, and you are right in pointing out a bias here. -- This address is for information only. I do not claim that these views are those of the Statistics Department or of Purdue University. Herman Rubin, Department of Statistics, Purdue University hrubin@stat.purdue.edu Phone: (765)494-6054 FAX: (765)494-0558 Start « Previous 1 2 3 4 5 6 Next » Related Discussion: XanGo (1 posts) these girls use this service for free - they just want phonesex from real guys 61 (1 posts) Eating Fish Protects Against Stroke (1 posts) Prevention of type 2 diabetes mellitus through inhibition of the Renin-Angiotensin system. (1 posts) Top Discussions From Our Archives Energy. Resilience. Vitality. Want some? Do you do website promotion with no success? Diabetes Spotted at an Earlier Age Other Discussion Categories Diabetes Symptoms Health Care Policy Health Care Providers HMO
|
| ||||||||
 |  |
