Great minds obviously think alike Helene, but then I did manage for
some strange reason to leave my computer at work last night . If
normal distributions are being mentioned then p = =0.01 on a standard
normal equates to 2.58 SEs, (or 2.62 if you want t on ~110df ... i
used qnorm and qt in R putting a p value of 0.005 since I assume they
reported a two sided test and i doubt it would make much difference
..0.01 would give you 2.3)
then if a difference of 5 is 2.58 standard errors then 5/2.68 is the
SE and the SD is that multipled by the square root of the sample size
(they seem pretty similar). That seems to give about 20, and I am not
sure I like th implications of that given your reported data, unless
teh distibutional assumption is wrong. Did they really do a t test> Or
did they may be do it on a log scale (process times are generally
skewed but often model well on a log normal distribution.. putting ln(5) in the
reverse calc gives 6.48 instead of 20 which seems far more sound, but
only if thats what they did..!
Andy
On 15 December 2015 at 17:03, Helene Hoegsbro Thygesen
<[log in to unmask]> wrote:
> Or you can assume that the p-value is 0.01 (or maybe a bit less than that?
> From published studies of distribution of p-values you can maybe infer the
> median p-value given that it is <0.01, but maybe something close to 0.01 is
> prudent since if it had been smaller than 0.001 they might have said so).
> And then you can find the SD that is consistent with the assumed p-value
>
> On Tue, Dec 15, 2015 at 4:56 PM, John Sorkin <[log in to unmask]>
> wrote:
>>
>> You might operate under the assumption that the mean + or - 3SDs includes
>> all your data, leading to the assumption that the range of your data is 6SD.
>> (40-11)/6
>> would thus be 1SD, (46-5)/6 would be the second SD.
>> John
>>
>> John David Sorkin M.D., Ph.D.
>> Professor of Medicine
>> Chief, Biostatistics and Informatics
>> University of Maryland School of Medicine Division of Gerontology and
>> Geriatric Medicine
>> Baltimore VA Medical Center
>> 10 North Greene Street
>> GRECC (BT/18/GR)
>> Baltimore, MD 21201-1524
>> (Phone) 410-605-7119
>> (Fax) 410-605-7913 (Please call phone number above prior to faxing)
>>
>> >>> David Epstein <[log in to unmask]> 12/15/15 9:50 AM >>>
>>
>> Dear Allstat
>>
>> I want to conduct meta-analysis comparing operating times of two
>> procedures across different studies.
>>
>> One paper gives the following in the table of results:
>>
>> Procedure A: Mean duration 24 minutes, range 11-40, N=108
>>
>> Procedure B: Mean duration 19 minutes, range 5-46, N=114
>>
>> P-value for difference <0.01, derived from unpaired t-test or Wilcoxon
>> test.
>>
>> To include this in a meta-analysis with results from other papers, I need
>> the Standard Deviation (I think?). Is there any way of recovering some
>> plausible estimate of SD from these results? Or, what assumptions would I
>> need to make to extract some estimate of SD?
>> Best wishes
>>
>> --
>> David Epstein
>> Departamento de Economía Aplicada
>> Universidad de Granada
>> Facultad de Ciencias Económicas y Empresariales
>> Campus de Cartuja, s/n, 18071. GRANADA
>> Teléfono:+34 958 249973 / 246258
>> FAX: +34 958 244046
>> email: [log in to unmask]
>> ==============================
>> Honorary Visiting Research Fellow
>> Centre For Health Economics
>> University of York
>> Heslington York Y010 5DD
>> email: [log in to unmask]
>> ==============================
>> You may leave the list at any time by sending the command
>>
>> SIGNOFF allstat
>>
>> to [log in to unmask], leaving the subject line blank.
>>
>>
>> Confidentiality Statement:
>>
>> This email message, including any attachments, is for the sole use of the
>> intended recipient(s) and may contain confidential and privileged
>> information. Any unauthorized use, disclosure or distribution is prohibited.
>> If you are not the intended recipient, please contact the sender by reply
>> email and destroy all copies of the original message.
>>
>> You may leave the list at any time by sending the command
>>
>> SIGNOFF allstat
>>
>> to [log in to unmask], leaving the subject line blank.
>
>
> You may leave the list at any time by sending the command
>
> SIGNOFF allstat
>
> to [log in to unmask], leaving the subject line blank.
You may leave the list at any time by sending the command
SIGNOFF allstat
to [log in to unmask], leaving the subject line blank.
|