Bayesian reanalysis using summary stats

PeterHoffenaar

Hi,

Given that the summary stats module can also handle F-statistics from ANOVA analyses, I would think that a t-statistic is no longer required to calculate the associated BFs. However, I noticed that in the literature a common approach seems to be to first calculate t-statistics from precise p-values and the combined sample size using Excel’s TINV function. The t-statistics are then used to calculate the BFs. So, my first question is whether it is better to use the ANOVA option in the summary stats module or to calculate the BF from a t-statistic.

Second, for a specific ANOVA the table in the published article mentions the F-statistic but degrees of freedom are not reported. Given this incomplete information, I could use the approach mentioned above and calculate the BF based on a t-statistic (as calculate from the p-value). Another option would be to asks the authors for the missing information on the degrees of freedom of the tests used. My question therefore is whether emailing the authors is worth the extra trouble?

Thanks for your help. Best,

Peter

EJ

Hi Peter,

"Given that the summary stats module can also handle F-statistics from ANOVA analyses" -- Can it?
" the ANOVA option in the summary stats module" -- but there isn't one.
Regardless, if you have the t-statistic, there is nothing wrong with applying the Summary Stat functionality. So even if we had an ANOVA Sum Stat option, there is nothing to gain.

If you have the ANOVA results, what you need is the number of subjects and the t-value for the comparison of interest. Sometimes you can make assumptions (equal variance), that, in concert with knowledge of means and standard deviations, allow you to obtain the relevant result. Otherwise, you have to ask the authors.

Cheers,
E.J.