Sample size for Repeated Measures

[ehsan sabaghian]

Dear all
I am working on a Cohort study with Repeated Measures. I want to calculate sample size. How do I
calculate the sample size?
Thanks in advance

[Pedro Emmanuel Alvarenga Americano do Brasil]

If you are looking for some simple e easy thing, I suggest this reading.

http://emj.bmj.com/cgi/content/full/22/3/180

You will need the figure one larger version to estimate minimum sample size.

Abraço forte e que a força esteja com você,

Dr. Pedro Emmanuel A. A. do Brasil
Instituto de Pesquisa Clínica Evandro Chagas
Fundação Oswaldo Cruz
Rio de Janeiro - Brasil

2009/11/16 ehsan sabaghian <e.sab...@gmail.com>

[Bruce Weaver]

You need to give more information. What is the outcome measure, for
example?

--
Bruce Weaver
bwe...@lakeheadu.ca
http://sites.google.com/a/lakeheadu.ca/bweaver/Home
"When all else fails, RTFM."

[ehsan sabaghian]

It's a continues variable.
I plan to measure it for five times:
1- before action
2- 3 days after
3- 14 days after
4- 28 days after
5- and 56 days after that.

[Pedro Emmanuel Alvarenga Americano do Brasil]

Sorry... I posted a answer for another email with similar heading. Please ignore it?

Abraço forte e que a força esteja com você,

Dr. Pedro Emmanuel A. A. do Brasil
Instituto de Pesquisa Clínica Evandro Chagas
Fundação Oswaldo Cruz
Rio de Janeiro - Brasil

2009/11/16 Pedro Emmanuel Alvarenga Americano do Brasil <emmanue...@gmail.com>

[Bruce Weaver]

> > On Nov 16, 8:34 am, ehsan sabaghian <e.sabagh...@gmail.com> wrote:
> > > Dear all
> > > I am working on a Cohort study with Repeated Measures. I want to
> > calculate
> > > sample size. How do I
> > > calculate the sample size?
> > > Thanks in advance
>
> > You need to give more information. What is the outcome measure, for
> > example?

On Nov 16, 10:06 am, ehsan sabaghian <e.sabagh...@gmail.com> wrote:
> It's a continues variable.
> I plan to measure it for five times:
> 1- before action
> 2- 3 days after
> 3- 14 days after
> 4- 28 days after
> 5- and 56 days after that.

OK, good. Now what is the effect you want to have power to detect,
and what is the minimum effect size that would be considered
practically important?

[ehsan sabaghian]

I want to compare 5 mean in one group in a period of time that I said. In a similar study with 3 time, they recorded 3 mean and standard deviation:

Time     Mean     SD
first       21.4        3
2nd       19.2       4.2
3rd        19.3       3.9

Now I want to do the same study with 5 time. The minimum effect is nearly between 2 to 4.

--
Ehsan Sabaghian
MSc Student of Biostatistics
Department of Biostatistics
Faculty of Public Health
Mashhad University of Medical Sciences (MUMS)

[Bruce Weaver]

On Nov 16, 3:11 pm, ehsan sabaghian <e.sabagh...@gmail.com> wrote:
> I want to compare 5 mean in one group in a period of time that I said. In a
> similar study with 3 time, they recorded 3 mean and standard deviation:
>
> Time     Mean     SD
> first       21.4        3
> 2nd       19.2       4.2
> 3rd        19.3       3.9
>
> Now I want to do the same study with 5 time. The minimum effect is nearly
> between 2 to 4.

If I follow, you have one group that will be measured 5 times, and you
want to have power to detect a mean change (drop) of at least 2 points
from time 1 to time 5, or at least to one of the time points after
time 1. Is this right?

What type of analysis do you intend to do? Repeated measures ANOVA?
Can the main question be boiled down to a paired t-test (e.g., time 1
vs. time 5)? If so, you could use Russ Lenth's sample size applet for
the paired t-test, available here:

http://www.cs.uiowa.edu/~rlenth/Power/

But you'll need an estimate of the SD of the change scores. To get
that, you need an estimate of the correlation between time points--
does that similar study report the correlations?

Compute the covariance from the correlation & the two standard
deviations:

COV(t1,t5) = r(t1,t5) * SD(t1) * SD(t5)

Then compute the variance of the change scores:

Var(change) = Var(t1) + Var(t5) - 2*COV(t1,t5)

And finally, take the square root of that to get the SD of the change
scores:

SD(change) = SQRT[Var(change)]

If you're not too confident in the estimate of the correlation, you
can try it with a range of plausible values. (This is often called a
"sensitivity analysis").

HTH.

[ehsan sabaghian]

Thank you for your advice, but I have to analyze data via Repeated Measures ANOVA method,
 so I have to estimate sample size for this kind of test.

[kornbrot]

G*Power does this – and more
http://www.psycho.uni-duesseldorf.de/aap/projects/gpower/
diana

On 17/11/2009 05:50, "ehsan sabaghian" <e.sab...@gmail.com> wrote:

Thank you for your advice, but I have to analyze data via Repeated Measures ANOVA method,
 so I have to estimate sample size for this kind of test.


On Tue, Nov 17, 2009 at 2:39 AM, Bruce Weaver <bwe...@lakeheadu.ca> wrote:

On Nov 16, 3:11 pm, ehsan sabaghian <e.sabagh...@gmail.com> wrote:
> I want to compare 5 mean in one group in a period of time that I said. In a
> similar study with 3 time, they recorded 3 mean and standard deviation:
>
> Time     Mean     SD
> first       21.4        3
> 2nd       19.2       4.2
> 3rd        19.3       3.9
>
> Now I want to do the same study with 5 time. The minimum effect is nearly
> between 2 to 4.

If I follow, you have one group that will be measured 5 times, and you
want to have power to detect a mean change (drop) of at least 2 points
from time 1 to time 5, or at least to one of the time points after
time 1.  Is this right?

What type of analysis do you intend to do?  Repeated measures ANOVA?
Can the main question be boiled down to a paired t-test (e.g., time 1
vs. time 5)?  If so, you could use Russ Lenth's sample size applet for
the paired t-test, available here:

   http://www.cs.uiowa.edu/~rlenth/Power/ <http://www.cs.uiowa.edu/%7Erlenth/Power/>

But you'll need an estimate of the SD of the change scores.  To get
that, you need an estimate of the correlation between time points--
does that similar study report the correlations?

Compute the covariance from the correlation & the two standard
deviations:

    COV(t1,t5) = r(t1,t5) * SD(t1) * SD(t5)

Then compute the variance of the change scores:

   Var(change) = Var(t1) + Var(t5) - 2*COV(t1,t5)

And finally, take the square root of that to get the SD of the change
scores:

   SD(change) = SQRT[Var(change)]

If you're not too confident in the estimate of the correlation, you
can try it with a range of plausible values.  (This is often called a
"sensitivity analysis").

HTH.

--
Bruce Weaver
bwe...@lakeheadu.ca
http://sites.google.com/a/lakeheadu.ca/bweaver/Home
"When all else fails, RTFM."

---

Professor Diana Kornbrot
  email:  d.e.ko...@herts.ac.uk    
   web:    http://web.mac.com/kornbrot/iweb/KornbrotHome.html
Work
School of Psychology
University of Hertfordshire
College Lane, Hatfield, Hertfordshire AL10 9AB, UK
    voice:     +44 (0) 170 728 4626
    mobile:   +44 (0) 796 890 2102
    fax          +44 (0) 170 728 5073
Home
19 Elmhurst Avenue
London N2 0LT, UK
   landline: +44 (0) 208 883 3657
   mobile:   +44 (0) 796 890 2102
   fax:         +44 (0) 870 706 4997

---