Your “best guess” is that the effect of a deworming treatment on happiness is a sudden benefit followed by a slow decline relative to no treatment? Do you have any theory of action that explains why this would be the case?
Trying to draw conclusions from such a dramatically underpowered study (with regard to this question) strikes me as absurd.
Our takeaway from this data is that there is not evidence of an effect (positive or negative).
We take these data to be our best guess because there are no prior studies of the effect of deworming on SWB, and the evidence of impact on other outcomes is very uncertain. However, all the effects are non-significant. We don’t have a theory of action because we think the overall evidence points to there being no effect (or at least just a very small one).
We ran the cost-effectiveness analysis as an exercise to see how deworming would look if we took the data at face-value. The point estimate was negative, but the confidence interval was so wide that the results were essentially uninformative, which converges with our conclusion that there is not a substantial effect of deworming on long-term wellbeing.
That being said, we can make assumptions that are favorable to deworming, such as assuming the effect cannot be negative. This, of course, involves overriding the data with prior beliefs — prior beliefs that we lack strong reasons to hold. In any case, we explore the results under these favorable assumptions in Appendix A2. In all plausible cases, deworming is still less cost-effective than StrongMinds, so even these exploratory analyses —which, again, we don’t endorse— don’t change our conclusion to not recommend deworming over StrongMinds.
Regarding power
Trying to draw conclusions from such a dramatically underpowered study (with regard to this question) strikes me as absurd.
It is unclear what evidence you use to claim the study is underpowered. As Joel mentioned in his comment to MichaelStJules (repasted below), we had 98% power to detect effect sizes of 0.08 SDs, the effect size that would make deworming more cost-effective than StrongMinds .
Assuming (generously) that the effects of deworming couldn’t become negative, then the initial would need to be 0.08 SDs for deworming to beat StrongMinds, which we had 98% power to detect.
In brief, we used Harrer et al.’s (2021) R library and their function for the power of a meta-analysis. With this, we computed the power we had to detect an initial effect (the intercept) of .08 SDs which was 97.88%. This uses the average sample sizes and the number of effect sizes.
We selected .08 as our smallest effect size of interest based on the results from our ‘optimistic’ model (see the Appendix A1)—which assumes the effect of deworming cannot be negative (not a model we endorse). The initial effect was .04 and the total cost-effectiveness was half that of StrongMinds. So, the effect would need to be twice as large to equal the cost-effectiveness of StrongMinds.
Your “best guess” is that the effect of a deworming treatment on happiness is a sudden benefit followed by a slow decline relative to no treatment? Do you have any theory of action that explains why this would be the case?
Trying to draw conclusions from such a dramatically underpowered study (with regard to this question) strikes me as absurd.
Hi Charles,
Our takeaway from this data is that there is not evidence of an effect (positive or negative).
We take these data to be our best guess because there are no prior studies of the effect of deworming on SWB, and the evidence of impact on other outcomes is very uncertain. However, all the effects are non-significant. We don’t have a theory of action because we think the overall evidence points to there being no effect (or at least just a very small one).
We ran the cost-effectiveness analysis as an exercise to see how deworming would look if we took the data at face-value. The point estimate was negative, but the confidence interval was so wide that the results were essentially uninformative, which converges with our conclusion that there is not a substantial effect of deworming on long-term wellbeing.
That being said, we can make assumptions that are favorable to deworming, such as assuming the effect cannot be negative. This, of course, involves overriding the data with prior beliefs — prior beliefs that we lack strong reasons to hold. In any case, we explore the results under these favorable assumptions in Appendix A2. In all plausible cases, deworming is still less cost-effective than StrongMinds, so even these exploratory analyses —which, again, we don’t endorse— don’t change our conclusion to not recommend deworming over StrongMinds.
Regarding power
It is unclear what evidence you use to claim the study is underpowered. As Joel mentioned in his comment to MichaelStJules (repasted below), we had 98% power to detect effect sizes of 0.08 SDs, the effect size that would make deworming more cost-effective than StrongMinds .
“we had 98% power to detect effect sizes of 0.08 SDs” How did you calculate this? This seems unlikley on a 3 point categorical scale like ths,
In brief, we used Harrer et al.’s (2021) R library and their function for the power of a meta-analysis. With this, we computed the power we had to detect an initial effect (the intercept) of .08 SDs which was 97.88%. This uses the average sample sizes and the number of effect sizes.
We selected .08 as our smallest effect size of interest based on the results from our ‘optimistic’ model (see the Appendix A1)—which assumes the effect of deworming cannot be negative (not a model we endorse). The initial effect was .04 and the total cost-effectiveness was half that of StrongMinds. So, the effect would need to be twice as large to equal the cost-effectiveness of StrongMinds.