sample size calculation

[Watson, Hunna]

Hello listers,

 

Anyone quick at calculating required sample sizes? Annoying government
worksite restrictions are blocking me from downloading power software
like G*Power.

I am after the required sample size for detecting a -large effect at 80%
power and alpha level of .05 for a 2x2x2 ANOVA with unequal cells. Can
anyone give me the required sample size and indicate which program they
used so I have a reference?

 

Regards,

Hunna

 

Hunna Watson

Research Officer

Centre for Clinical Interventions (CCI)

Department of Health in Western Australia

 

223 James St

Northbridge WA 6003

Tel: [phone]

Fax: [phone]

 

www.cci.health.wa.gov.au

 

[Marta Garcia-Granero]

Hi Hunna

MorePower (see ref. below) is a great freeware utility for ANOVA designs. Unfortunately, you don't give all the required details

- Which effect are you interested in? (main effects, 2-way interactions, three-way interactions...)
- You mention unequal cells. How big the unbalance is? The sample size needed can increase dramatically if the design is heavily unbalanced.

Assuming that when you indicate "large effect" you are talking about a partial eta-square of 0.14, MorePower output syas (for a perfectly balanced design) that each cell should have at least 8 cases (since a 2x2x2 design has 8 cells, the total sample size required is: Nt=64). This is the sample size needed for the 3-way interaction. Figures for main effects and 2-way interactions were slightly lower (total sample sizes needed Nt=60 and Nt=56, respectively).

MorePower doesn't compute sample sizes for unbalanced designs. If you are going to have unequal cells, make sure that the harmonic mean of your sample sizes is equal to the one given by the program (for each cell).

Anyway: is a large effect expected (based on your previous experience) or should you consider the possibility of finding smaller effects?

Program reference:
"MorePower by [hidden email],
Dept. of Psychology, University of Saskatchewan, October 31, 2006.
Implementation documentation and details about the built-in examples at: <a class="moz-txt-link-freetext" href="http://">http://
duke.usask.ca/~campbelj/work/MorePower.html
If you use the calculator, please cite Campbell, J.I.D. & Thompson, V.A. (2002). More power to you: Simple power calculations for treatment effects with one degree of freedom. Behavior Research Methods, Instruments, and Computers, 34, 332-337."

HTH,
Marta Garcia-Granero

Anyone quick at calculating required sample sizes? Annoying government
worksite restrictions are blocking me from downloading power software
like G*Power.

I am after the required sample size for detecting a -large effect at 80%
power and alpha level of .05 for a 2x2x2 ANOVA with unequal cells. Can
anyone give me the required sample size and indicate which program they
used so I have a reference?

[Swank, Paul R]

Most power software does not handle unequal cell sizes but this can
significantly effect the power calculations. I usually do simulations to
check power for such designs. Not only does it not require outside
programs, it also forces you to undertsand you model and your data more
thoroughly in order to do the simulation. The hardest part is to decide
on the effect sizes. Cohen's examples are mow often seen as too large
for many kinds of studies so you need to be aware of what typical effct
sizes are for your research area.

Paul R. Swank, Ph.D. Professor
Director of Reseach
Children's Learning Institute
University of Texas Health Science Center-Houston


-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Watson, Hunna
Sent: Wednesday, September 19, 2007 12:10 AM
To: [hidden email]
Subject: sample size calculation

Hello listers,



Anyone quick at calculating required sample sizes? Annoying government
worksite restrictions are blocking me from downloading power software
like G*Power.

I am after the required sample size for detecting a -large effect at 80%
power and alpha level of .05 for a 2x2x2 ANOVA with unequal cells. Can
anyone give me the required sample size and indicate which program they
used so I have a reference?



Regards,

Hunna



Hunna Watson

Research Officer

Centre for Clinical Interventions (CCI)

Department of Health in Western Australia



223 James St

Northbridge WA 6003

Tel: [phone]

Fax: [phone]



www.cci.health.wa.gov.au

[zstatman]

The link for MorePower is now
http://homepage.usask.ca/~jic956/work/MorePower.html

WMB

  _____

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Marta Garcia-Granero
Sent: Wednesday, September 19, 2007 12:09 PM
To: [hidden email]
Subject: Re: sample size calculation


Hi Hunna

MorePower (see ref. below) is a great freeware utility for ANOVA designs.
Unfortunately, you don't give all the required details

- Which effect are you interested in? (main effects, 2-way interactions,
three-way interactions...)
- You mention unequal cells. How big the unbalance is? The sample size
needed can increase dramatically if the design is heavily unbalanced.

Assuming that when you indicate "large effect" you are talking about a
partial eta-square of 0.14, MorePower output syas (for a perfectly balanced
design) that each cell should have at least 8 cases (since a 2x2x2 design
has 8 cells, the total sample size required is: Nt=64). This is the sample
size needed for the 3-way interaction. Figures for main effects and 2-way
interactions were slightly lower (total sample sizes needed Nt=60 and Nt=56,
respectively).

MorePower doesn't compute sample sizes for unbalanced designs. If you are
going to have unequal cells, make sure that the harmonic mean of your sample
sizes is equal to the one given by the program (for each cell).

Anyway: is a large effect expected (based on your previous experience) or
should you consider the possibility of finding smaller effects?

Program reference:
"MorePower by [hidden email],
Dept. of Psychology, University of Saskatchewan, October 31, 2006.
Implementation documentation and details about the built-in examples at:
http://
duke.usask.ca/~campbelj/work/MorePower.html
If you use the calculator, please cite Campbell, J.I.D. & Thompson, V.A.
(2002). More power to you: Simple power calculations for treatment effects
with one degree of freedom. Behavior Research Methods, Instruments, and
Computers, 34, 332-337."

HTH,
Marta Garcia-Granero



Anyone quick at calculating required sample sizes? Annoying government

worksite restrictions are blocking me from downloading power software

like G*Power.



I am after the required sample size for detecting a -large effect at 80%

power and alpha level of .05 for a 2x2x2 ANOVA with unequal cells. Can

anyone give me the required sample size and indicate which program they

used so I have a reference?

[Kornbrot, Diana]

Usually, the analysis is only as good as the weakest link
So, I estimate power assuming that all groups have the same N as the
smallest group
This is conservative, the power will be at least as great as that obtained
by this method
It emphasise that insituations where one group is hard to recruit, for
whatever reason, eg. Clinical samples, uppung the number in the larger
groups does not help much
Also, wherever possible, determine the real magnitude you want to detect
e.g. Weight loss in kg, rahter than an arbitrary statistical effect size

Best

diana


On 19/9/07 17:16, "Swank, Paul R" <[hidden email]> wrote:

> Most power software does not handle unequal cell sizes but this can
> significantly effect the power calculations. I usually do simulations to
> check power for such designs. Not only does it not require outside
> programs, it also forces you to undertsand you model and your data more
> thoroughly in order to do the simulation. The hardest part is to decide
> on the effect sizes. Cohen's examples are mow often seen as too large
> for many kinds of studies so you need to be aware of what typical effct
> sizes are for your research area.
>
> Paul R. Swank, Ph.D. Professor
> Director of Reseach
> Children's Learning Institute
> University of Texas Health Science Center-Houston
>
>
> -----Original Message-----
> From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
> Watson, Hunna
> Sent: Wednesday, September 19, 2007 12:10 AM
> To: [hidden email]
> Subject: sample size calculation
>
> Hello listers,
>
>
>
> Anyone quick at calculating required sample sizes? Annoying government
> worksite restrictions are blocking me from downloading power software
> like G*Power.
>
> I am after the required sample size for detecting a -large effect at 80%
> power and alpha level of .05 for a 2x2x2 ANOVA with unequal cells. Can
> anyone give me the required sample size and indicate which program they
> used so I have a reference?
>
>
>
> Regards,
>
> Hunna
>
>
>
> Hunna Watson
>
> Research Officer
>
> Centre for Clinical Interventions (CCI)
>
> Department of Health in Western Australia
>
>
>
> 223 James St
>
> Northbridge WA 6003
>
> Tel: [phone]
>
> Fax: [phone]
>
>
>
> www.cci.health.wa.gov.au
>

Professor Diana Kornbrot
Evaluation Co-ordinator, Blended Learning Unit
University of Hertfordshire
College Lane, Hatfield, Hertfordshire AL10 9AB, UK

email:  [hidden email]
web:    http://web.mac.com/kornbrot/iweb/KornbrotHome.html
         
Blended Learning Unit
    voice [phone]
    fax    [phone]
Psychology        
    voice [phone]
    fax    [phone]
Home
19 Elmhurst Avenue
London N2 0LT, UK
   voice [phone]

[cbautista]

Hi list,

 

As far as I know, with SPSS only a histogram can be displayed, so I was wondering if someone has a script to plot two or more histograms. I have four variables, and I'd would like to display the four distributions at the same picture (or at least two). Maybe, other statistical package can do it.

 

thanks in advance,

 

/Christian

[ViAnn Beadle]

What version of SPSS do you have and how do you want the histograms
arranged? You can do this with GPL in SPSS 15 (maybe 14 as well but I don't
have a way to test it)



From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Christian Bautista
Sent: Saturday, September 29, 2007 8:46 AM
To: [hidden email]
Subject: hISTROGRAMS....



Hi list,



As far as I know, with SPSS only a histogram can be displayed, so I was
wondering if someone has a script to plot two or more histograms. I have
four variables, and I'd would like to display the four distributions at the
same picture (or at least two). Maybe, other statistical package can do it.



thanks in advance,



/Christian

=

[John M. Trent]

I am sure this is more simple than what you are looking for however

EXAMINE

VARIABLES=(INSERT YOUR VARIABLE NAMES HERE)

/PLOT HISTOGRAM

/STATISTICS NONE

/CINTERVAL 95

/MISSING LISTWISE

/NOTOTAL.

This will produce as many histograms as you have variables...I simply cut and paste into WORD for comparative purposes...but you may want something I don't quite understand...



John Mark Trent, PhD
Educational Psychologist
Family Therapist

Change occurs only as we begin thinking about and working on the self ---
rather than staying focused on and reactive to the other
http://relationalgrace.blogspot.com/



________________________________

From: SPSSX(r) Discussion on behalf of Christian Bautista
Sent: Sat 9/29/2007 9:45 AM
To: [hidden email]
Subject: hISTROGRAMS....


Hi list,

As far as I know, with SPSS only a histogram can be displayed, so I was wondering if someone has a script to plot two or more histograms. I have four variables, and I'd would like to display the four distributions at the same picture (or at least two). Maybe, other statistical package can do it.

thanks in advance,

/Christian