On Jul 20, 4:19 am, James100 <marylan...@[EMAIL PROTECTED]
> wrote:
> On Jul 16, 9:40 am, Ray Koopman <koop...@[EMAIL PROTECTED]
> wrote:
>> On Jul 15, 9:31 pm, James100 <marylan...@[EMAIL PROTECTED]
> wrote:
>>>>
>>>> SEM = ??? Standard Error of a Mean?
>>>> That would make sense only if all the groups were the same size.
>>>> What were the n's?
>>>
>>> Many thanks for the response. n's were 3 for all groups.
>>
>> If SEM is the Standard Error of a Mean, and each mean is based on
>> 3 observations, then I assume that SEM was computed as sqrt(MSE/3),
>> where MSE is the pooled Mean Square for Error from the anova. With
>> only 3 observations/group, you're probably better off using the pooled
>> error term, which averages the variances of all six groups, than you
>> would be if you were to use only the variances of the two groups whose
>> means you're comparing, unless the true variances of those two groups
>> differ substantially from the true variances of the four other groups.
>> Notice that this refers to the true variances, not the sample
>> variances. Sample variances based on only 3 values are very poor
>> estimates of the true variances, so even if you had the sample
>> variance for each group, I would still recommmend that you use the
>> average of all six. This is a case where you need to be willing to
>> accept the possibility of some bias in your estimate of the variance
>> in exchange for an almost certain reduction in the standard error of
>> that estimate.
>> Mean(Trt) - Mean(Control)
>> Anyhow, your effect size estimate is -------------------------.
>> SEM * sqrt(3)
>
> Dear Ray,
>
> Have send a message to you directly, but don't know if you have
> received it? MANY many thanks fot the very usefull help.
>
> James
Yes, I got it. You're quite welcome. Since you have no local sup****t,
feel free to ask here, even for the little things.
|
