Sample size, timing, and other confounding factors: toward a fair assessment of stay-at home orders.

Besançon, Lonni; Meyerowitz-Katz, Gideon; Flahault, Antoine

The small sample size (n=10) and the sample's composition have not been justified and introduce a lack of representativeness. Brauner et al. [2] (n=41), along with others, present contradictory results with bigger samples that are not discussed in the manuscript. The authors do not present adequate reasoning for such a small pool of data and exclude numerous countries that provide similarly appropriate data (e.g., Switzerland by canton). Given the small sample size, the study is also at a very high risk of uncontrolled confounding. For instance, Sweden and Iran are directly compared in terms of the outcomes of their interventions without considering the plethora of cultural, social, and political differences between them that might impact their case count. This is worrisome, because there are examples of other, more culturally similar, nations which could have been included in either group, such as Denmark for which case counts by administrative region are publicly available online. We therefore argue that the potential for unmeasured confounding could have been reduced with a larger and more careful composition of the sample and should be directly acknowledged.