sample size for simple mediation

[Salbod]

Good Afternoon (NY),

 

Preacher and Hayes (2008a)  suggest that  “bootstrapping enables researchers to use smaller samples than would be necessary to satisfy the distributional assumptions of other methods (although samples should not be too small).”

 

My question is how small a sample can bootstrapping be applied to. I’ve used Preacher and Hayes (2008b) SPSS Macro  for Multiple Mediation for a series of simple mediation hypotheses. The samples varied in size from 16 to 29. Is there a minimal sample size mediation bootstrapping? Kind of like a rule-of-thumb.

 

Any suggestions or references will be greatly appreciated.

 

Stephen Salbod, Pace University, NYC

 

 

 

 

 

 

Preacher, K.J., & Hayes, A.F. (2008a). Contemporary approaches to assessing mediation in communication research. In A. F. Hayes, M. D. Slater, & L. B. Snyder(Eds.), The Sage sourcebook of advanced data analysis methods for communication research (pp. 13-54). Thousand Oaks, CA: Sage.

 

Preacher, K.J., & Hayes, A.F. (2008b). Asymptotic and resampling strategies for assessing and comparing indirect effects in multiple mediator models. Behavior Research Methods, 40, 879-891.

 

 

[Stephen J. Toglia]

Here you go:

Hayes response to me "Mathematically, the lower limit would be the sample size minus one more than the total number of predictor variables in the model you are estimating."

I hope this help.

Stephen

--- On Sat, 9/11/10, Salbod, Mr. Stephen <[hidden email]> wrote:

From: Salbod, Mr. Stephen <[hidden email]>
Subject: sample size for simple mediation
To: [hidden email]
Date: Saturday, September 11, 2010, 6:05 PM

Good Afternoon (NY),

 

Preacher and Hayes (2008a)  suggest that  “bootstrapping enables researchers to use smaller samples than would be necessary to satisfy the distributional assumptions of other methods (although samples should not be too small).”

 

My question is how small a sample can bootstrapping be applied to. I’ve used Preacher and Hayes (2008b) SPSS Macro  for Multiple Mediation for a series of simple mediation hypotheses. The samples varied in size from 16 to 29. Is there a minimal sample size mediation bootstrapping? Kind of like a rule-of-thumb.

 

Any suggestions or references will be greatly appreciated.

 

Stephen Salbod, Pace University, NYC

 

 

 

 

 

 

Preacher, K.J., & Hayes, A.F. (2008a). Contemporary approaches to assessing mediation in communication research. In A. F. Hayes, M. D. Slater, & L. B. Snyder(Eds.), The Sage sourcebook of advanced data analysis methods for communication research (pp. 13-54). Thousand Oaks, CA: Sage.

 

Preacher, K.J., & Hayes, A.F. (2008b). Asymptotic and resampling strategies for assessing and comparing indirect effects in multiple mediator models. Behavior Research Methods, 40, 879-891.

 

 

[Salbod]

 

 

From: Salbod, Mr. Stephen
Sent: Tuesday, September 21, 2010 6:29 PM
To: 'Stephen J. Toglia'
Subject: RE: sample size for simple mediation

 

Hi Stephen,

 

Thanks for the rule-of-thumb. I had the student I’m working with email Preacher about the sample size in her study. Here is his reply:

 

“I wish it were true that bootstrapping could solve power problems in small samples, but unfortunately it only helps a little. If you have Ns that low you will still probably have low power, but it is also true that as N decreases, the *relative* benefit of bootstrapping over normal-theory methods increases.

N = 16 is not technically too small to use bootstrapping, but with really small Ns you can often run into estimation problems because the occasional bootstrap resample might generate a constant for one of the variables. This is particularly likely to happen when you have at least one binary variable, e.g. gender, such that the occasional resample could have all 0s or all 1s for gender. This causes regression to fail.”

 

Regards, Steve

 

From: Stephen J. Toglia [mailto:[hidden email]]
Sent: Tuesday, September 21, 2010 4:21 PM
To: Salbod, Mr. Stephen; [hidden email]
Subject: Re: sample size for simple mediation

 

Here you go:

Hayes response to me "Mathematically, the lower limit would be the sample size minus one more than the total number of predictor variables in the model you are estimating."

I hope this help.

Stephen

 

 

--- On Sat, 9/11/10, Salbod, Mr. Stephen <[hidden email]> wrote:

From: Salbod, Mr. Stephen <[hidden email]>
Subject: sample size for simple mediation
To: [hidden email]
Date: Saturday, September 11, 2010, 6:05 PM

Good Afternoon (NY),

 

Preacher and Hayes (2008a)  suggest that  “bootstrapping enables researchers to use smaller samples than would be necessary to satisfy the distributional assumptions of other methods (although samples should not be too small).”

 

My question is how small a sample can bootstrapping be applied to. I’ve used Preacher and Hayes (2008b) SPSS Macro  for Multiple Mediation for a series of simple mediation hypotheses. The samples varied in size from 16 to 29. Is there a minimal sample size mediation bootstrapping? Kind of like a rule-of-thumb.

 

Any suggestions or references will be greatly appreciated.

 

Stephen Salbod, Pace University, NYC

 

 

 

 

 

 

Preacher, K.J., & Hayes, A.F. (2008a). Contemporary approaches to assessing mediation in communication research. In A. F. Hayes, M. D. Slater, & L. B. Snyder(Eds.), The Sage sourcebook of advanced data analysis methods for communication research (pp. 13-54). Thousand Oaks, CA: Sage.

 

Preacher, K.J., & Hayes, A.F. (2008b). Asymptotic and resampling strategies for assessing and comparing indirect effects in multiple mediator models. Behavior Research Methods, 40, 879-891.