[ABE-L] Fwd: {MEDSTATS} Re: "Basic and Applied Psychology" Bans The Use of P Values and Confidence Intervals

[Basilisk De Braganca Pereira]

Enviado do meu iPhone

Início da mensagem encaminhada

> De: Basilio De Braganca Pereira <basiliopereira em gmail.com>
> Data: 1 de março de 2015 17:38:50 BRT
> Para: basilio em hucff.ufrj.br
> Assunto: Enc.: {MEDSTATS} Re: "Basic and Applied Psychology" Bans The Use of P Values and Confidence Intervals
> 
> 
> 
> Enviado do meu iPhone
> 
> Início da mensagem encaminhada
> 
>> De: roland andersson <rolandersson em gmail.com>
>> Data: 1 de março de 2015 16:23:11 BRT
>> Para: medstats <medstats em googlegroups.com>
>> Assunto: Re: {MEDSTATS} Re: "Basic and Applied Psychology" Bans The Use of P Values and Confidence Intervals
>> Responder A: medstats em googlegroups.com
>> 
>> As a non-statistician I have followed this discussion with interest. 
>> 
>> One may make inferences about some results in many ways and get slightly different results - be it p-values or confidence intervals. To me as a clinician the effect measure will be more important than the p-value. A large effect with a p-value <0.05 will give a stronger impression to me than a small effect with p<0.0001. 
>> 
>> But can one single study claim to reflect the truth? And does one statistical method describe the truth more truely than the other? And how important is the difference between the one and the other? To me one single study does not prove anything. It can only say something about that specific sample that was studied and can not be generalised irrespective of the the size of the p-value or CI with or without adjustments for multiple comparisons. I can never trust that single result when I will make decisions for my next patient. I need more studies that shows similar result. 
>> 
>> But even then there will always remain some uncertainty. We study human beings and the variability between individuals are much larger than any confidence interval can show. And the measurements are imperfect. This must also be taken into account when we judge the results. These errors does not disappear no matter how exact mathematical methods we use. I like to use the theoretically most accurate statistical method, but must still be humble in the interpretation because of these errors. 
>> 
>> Roland Andersson
>> Surgeon
>> 
>> 
>> 
>> 
>> 
>> 
>> 
>> 
>> 
>> 2015-02-27 21:22 GMT+01:00 John Whittington <John.W em mediscience.co.uk>:
>>> At 08:56 28/02/2015 +1300, Paul Barrett wrote:
>>>> Hello John .... Let me quote from Bryan Manly, ....
>>>> "....Randomization testing is a way of determining whether the null hypothesis is reasonable in this type of situation. A statistic S is chosen to measure the extent to which data show the pattern in question. The value s of S for the observed data is then compared with the distribution of S that is obtained by randomly reordering the data. The argument made is that if the null hypothesis is true, then all possible orders for the data were equally likely to have occurred. .... The significance level of s is the proportion or percentage of values that are as extreme or more extreme than this value in the randomization distribution. This can be interpreted in the same way as for conventional tests of significance...
>>> 
>>> As you can see, what is essentially being described is simply a distribution-independent 'NHST'
>>> 
>>>> The important bit (for me) are those last three sentences. ... Although I had never considered randomization as invoking a formal null-hypothesis as might be set out using conventional inferential methods (invoking random sampling from some defined hypothetical population), I think the above suggests that a ‘null hypothesis’ is nevertheless being invoked, albeit quite different to that encountered in data-model oriented inference.
>>>> So I stand corrected.
>>> 
>>> Indeed. Re-sampling methods have the advantage of making no assumptions about distribution (although there obviously are plenty of hypothesis tests {aka 'non-parametric tests'} which are largely distribution-independent), but they are still essentially just a way of undertaking a 'NHST', since the re-sampling process itself is, in effect, creating a 'null hypothesis'.  Such procedures certainly do (or can) produce a 'p-value' - whether one calls is that or not!
>>> 
>>> 
>>> Kind Regards, 
>>> 
>>> 
>>> 
>>> John
>>> 
>>> ----------------------------------------------------------------
>>> Dr John Whittington,       Voice:    +44 (0) 1296 730225
>>> Mediscience Services       Fax:      +44 (0) 1296 738893
>>> Twyford Manor, Twyford,    E-mail:   John.W em mediscience.co.uk
>>> Buckingham  MK18 4EL, UK             
>>> ----------------------------------------------------------------
>>> -- 
>>> -- 
>>> To post a new thread to MedStats, send email to MedStats em googlegroups.com .
>>> MedStats' home page is http://groups.google.com/group/MedStats .
>>> Rules: http://groups.google.com/group/MedStats/web/medstats-rules
>>> 
>>> --- 
>>> You received this message because you are subscribed to the Google Groups "MedStats" group.
>>> To unsubscribe from this group and stop receiving emails from it, send an email to medstats+unsubscribe em googlegroups.com.
>>> For more options, visit https://groups.google.com/d/optout.
>> 
>> -- 
>> -- 
>> To post a new thread to MedStats, send email to MedStats em googlegroups.com .
>> MedStats' home page is http://groups.google.com/group/MedStats .
>> Rules: http://groups.google.com/group/MedStats/web/medstats-rules
>> 
>> --- 
>> You received this message because you are subscribed to the Google Groups "MedStats" group.
>> To unsubscribe from this group and stop receiving emails from it, send an email to medstats+unsubscribe em googlegroups.com.
>> For more options, visit https://groups.google.com/d/optout.
-------------- Próxima Parte ----------
Um anexo em HTML foi limpo...
URL: <https://lists.ime.usp.br/archives/abe/attachments/20150301/36d4b6c8/attachment.html>