Baumsteiger (2019) used a pre-post design with a control group. The reported means show that there was a significant pre-intervention difference between the two groups. Therefore, to calculate the effect size and its standard error, we followed a procedure proposed by Klauer (2001). That is, we separately calculated the between-group effect sizes before and after the intervention and then subtracted them from each other. Those calculation were performed in Matlab: [hedges_g_pre,SE_g_pre] = calculateEffectSize(4.08,4.62,1.28,1.41,43,32) [hedges_g_post,SE_g_post] = calculateEffectSize(4.8,4.84,1.41,1.49,43,32) effect_size_estimate_baumsteiger = hedges_g_post - hedges_g_pre Because pre-test and post-test scores are correlated (rho_bs=0.47), the standard error of this effect size estimate is roughly sqrt(SE_g_pre^2+SE_g_post^2-2*rho*SE_g_pre*SE_g_post), where SE_g_pre and SE_g_post are the standard errors of Hedge's g for the comparisons between the two groups before the intervention and after the intervention, respectively.

The standard error is only a rough estimate. I am planning to replace it using the more accurate formula from Equation 25 in Morris (2008).

This could be improved by using the d_PPC2 effect size estimator instead of Hedges g.

The estimate of the half-life of positive emotion was not obtained with the new method. Moreover, 26 minutes is currently used as a point estimate. So the uncertainty about its value is not yet taken into account. This can be improved.

The estimate of the half-life of positive emotion was not obtained with the new method. Moreover, 26 minutes is currently used as a point estimate. So the uncertainty about its value is not yet taken into account. This can be improved.

What is the units of the effect size?

26 minutes is the median duration of an emotional episode of joy in Study 1. That is not the same as the half-life. Under the assumption of exponential decay, the half-life would be shorter because the emotional episode doesn't end at 50% intensity. This has to be corrected. I am currently working on it. In Study 2, the median duration of joy was 19 minutes, and the median duration of gratitude was 13 minutes.

The current model assumes that the effect size is for the first day after the intervention, but it is actually for the first month. The model needs to be adjusted to capture this.

Is there an assumption that Morris (2008) makes in order to control for pre-existing differences in treatment and control groups?

The references for the meta-analyses have yet to be added.

to do: discuss strengths of the study to do: acknowledge that the study wasn't preregistered to do: explicitly say that we need a larger RCT? to do: add information about how people can get started with using Squiggle

Maybe it's just me, but I like hyperlinks for citations so we can work out what paper you are talking about more easily. Same with WELLBYs #WeWantLinks I'm sorry about the performance issues. We don't have this problem in non-observable environments! I've spent some time trying to work it out but haven't had much luck. It's definitely a priority for me to fix.

As per comment in your notebook, the stylesheet would be great to import here. Makes it easier for the eyes

This footnote doesn't work. It should be #footnote1

Also, our playground is a good place to play with Squiggle: https://www.squiggle-language.com/playground/

I'm loving this way of sectioning off things.

Against Malaria Foundation

Squiggle is a (close to) pure language, is there any reason why you have days_per_month() and not just days_per_month?