A priori power analysis using G*Power: Estimating required sample size for multiple regression
HTML-код
- Опубликовано: 2 май 2020
- This video demonstrates how to perform power analyses to arrive at sample size projections for tests of the multiple R-square and an individual regression slope using the G*Power program (www.psychologie.hhu.de/arbeit.... Download a copy of the Powerpoint referenced in the video here: drive.google.com/open?id=1Pqt...
Indeed these contents are useful for all types of researchers. Many thanks to you Mike.
Thank you so much for this! Has truly saved my sanity as I work on my dissertation
Thank you!! Stats is the worst but this makes it so much easier to figure out rather than just clicking everything
Great Effort! Thank you !
Thank you very much!!
Thank you for GREAT videos! Wish I could take a series of your stats classes. Your HLM videos really helped me with my last dissertation project! Defending in late June - this Summer.. via Zoom!
Hi Carolyn, I'm so glad you have found my videos helpful! Good luck with your dissertation defense this summer. I'm rooting for you!
Hi @@mikecrowson2462 Thank you very much for the helpful videos you post on this platform. i am currently conducting an experimental study and my suppervisor told me to use G*power to estimate the sample size so my question is how can i assume the R-square. i mean i dont understand how you get the R-square so kindly can you help me. i will be greatful!
@@abdishakurdiriye4673 Hi there. Thanks for visiting. The R-square is your projection regarding the population R-square. You will never 'know' what it is (and if you did, then why do the study, right? haha). So think of it as your best 'guess'/'estimate' given what you currently understand from your literature in your field (where there are previous studies that may lead you to come up with a plausible value; and/or where there is theory that you draw from or construct that might lead you to speculate about the possible R-square). If there is literally nothing out there to help you make an educated guess, then I suppose you could pick the smallest R-square (effect) you think would be treated as meaningful in your field and then use that as a basis for estimation. I believe G*power defaults with Cohen's value for a 'medium' effect. You can hover your cursor over the R-square and you should see what they would be for small and large effects. Hope this helps.
Thanks!
Great video that has really helped me out. Am I correct in assuming that having completed both levels, you would opt for the larger of the two stated sample sizes, or would you need to recruit the sum total of the two figures?
Thank you. It is great video beneficial for alll. Can you please help to understand whether ex-ante power analysis and a priori power analysis the same thing or different?
helllo mister! im very new to this thank you very much!
For a priori power analysis, do you know if we need to do it for topic modelling LDA research too?
Very useful video. If my model includes three independent variables and one moderator, how many predictors should I specify? Thank you sir
a great video sir. can i get the reference for the notes for the first explanation about g*power? tq in advanced
Please tell what to do if i have three groups? how many samples to be taken in each group?
Can I ask why we change it from .30 to .60. I'm trying to work out sample size with 4 predictors so assume r2 is .40 so would mine then change to .80?
Where did he get .30 from?
Mr. Crowson. I have been using G Power for some time now and am hoping I can email you a question I have about A Priori. Reply to this comment so we may connect. Thank you.
Is the population R2 of 0.30 is applicable for unknown population?
Hi Aidza, the question of whether R-square=.30 is applicable in your study can only be made based on either (a) previous research where similar R-square values were obtained or (b) an expectation based on theory of a population effect that large. If you have reason to suspect a population R-square that large (from a and/or b above), then you can go with it. If you don't have any theoretical or empirical basis for projecting a given population R-square, then the question of whether .30 is applicable really does not make sense. An R-square of .30 would be considered a large effect, according to Cohen's thresholds (see powerpoint). However, the question is whether it is reasonable depends on whether there is reason to expect such a large population effect - and this takes you back to theory and previous research. If you are expecting a large effect by proposing R-square =. .30 and the actual effect is less, then your sample size projection will likely produce test results (based on your sample) with lower power than you desired. Personally, I would rather underestimate the population effect rather than overestimate it when performing power analysis to arrive at a sample size projection - although I also don't want the understimation to be too great either. One possibility might be to create sample size projections for a range of R-square values and then make your decision after considering various research scenarios involving different population estimates for R-square. But I can't stress it enough. Make your decision about what R-square you need based on consideration of theory and prior research in your field. I know this is a long-winded answer, but the issues raised are complex :) I hope this is helpful to you. Cheers!
@@mikecrowson2462 Mr Mike Crowson, thank you for replying my comment. Due to extend my further understanding, I have email you to ask your guidance regarding G-Power with Multiple regression.