Hi Gemma, I know it sounds odd to enter a mean for an SD! But when you model the predictions of the alternative hypothesis with a normal centered on zero (i.e. zero population difference in means between conditions), or with a half-normal with the highest point being over zero (i.e. zero population difference in means between conditions), then the SD of that normal (or half-normal) scales the distribution with the sort of effects you find plausible. It means you think the population difference for Study 2 is most likely between 0 and Study 1's mean difference (with a probability of a third that the true population mean difference for Study 2 is greater than Study 1's mean difference). The short answer is: You enter as the SD the sort of effect that is plausible; and Study 1's mean difference provides you with an objective estimate of the sort of effect that is plausible. Hope that makes sense!
Use this: Dienes, Z. (2014). Using Bayes to get the most out of non-significant results. Frontiers in Psycholology, 5: 781. doi: 10.3389/fpsyg.2014.00781
Hi Zoltan, for whether the distribution is 1 or 2 tailed you said that as your data is half normal you should enter a value of 1. How do you know if your data is half normal?
Hi Gemma, I know it sounds odd to enter a mean for an SD! But when you model the predictions of the alternative hypothesis with a normal centered on zero (i.e. zero population difference in means between conditions), or with a half-normal with the highest point being over zero (i.e. zero population difference in means between conditions), then the SD of that normal (or half-normal) scales the distribution with the sort of effects you find plausible. It means you think the population difference for Study 2 is most likely between 0 and Study 1's mean difference (with a probability of a third that the true population mean difference for Study 2 is greater than Study 1's mean difference). The short answer is: You enter as the SD the sort of effect that is plausible; and Study 1's mean difference provides you with an objective estimate of the sort of effect that is plausible. Hope that makes sense!
Use this: Dienes, Z. (2014). Using Bayes to get the most out of non-significant results. Frontiers in Psycholology, 5: 781. doi: 10.3389/fpsyg.2014.00781
Very useful thanks very much for making this available.
Hi Zoltan, for whether the distribution is 1 or 2 tailed you said that as your data is half normal you should enter a value of 1. How do you know if your data is half normal?
Hi Zoltan,
How come you enter Study 1's mean difference in the standard deviation box in the video?
Thanks
Gemma
The SD scales the sort of size of effect you expect - and the mean for one study informs you what to expect for the next. Hope that helps!
Greetings