 |
 |
Shau-Ming, The reviewer is suggesting two different things that need to be done. First, they are asking for a "sensitivity analysis" and second they are asking if the experiment is properly powered. Power computations of an already conducted study are not valid. You are supposed to compute the power prior to the experiment. I would simply state that post-hoc power calculations aren't valid. If you still want to compute the power, you can use G*Power as suggested previously. An alternative would be to compute the power based on previous experimental results and say that you had X% power based on the literature. The "sensitivity analysis" is an analysis that looks at how the effect changes when you change the variables in the model. If changing the variables in your model impacts the results, then the results are very sensitive to the covariates in the model. Best Regards, Donald McLaren, PhD On Mon, Oct 17, 2016 at 1:28 PM, JLabus <[log in to unmask]> wrote: > I believe this is an option in G*Power > > > > On 10/17/2016 7:29 AM, Wei, Shau-Ming (NIH/NIMH) [F] wrote: > >> Dear SPM users, >> >> One reviewer suggested that we perform a "sensitivity analysis" to detect >> sufficient power for the results. I am not aware of a sensitivity analysis >> that can determine the power needed for a study. Is it the same as a >> Cohen's power analysis or is this something else? >> >> Thanks, >> Shau-Ming Wei >> > > -- > Jen Labus, PhD > Associate Professor > Director, Bioinformatics and Neuroimaging Core > G. Oppenheimer Center for Neurobiology of Stress and Resilience > David Geffen School of Medicine at UCLA > CHS 42-249 MC 737818 > 10833 Le Conte Avenue > Los Angeles, CA 90095-7378 > phone: 310.206.0738*** fax: 310.825.1919 > e.mail: [log in to unmask] > CNS website: http://painrepository.org >