Working through lockdown.

wallywombat:

RB1981:

Indeed the recent Damask-19 research in Denmark found no particular evidence for the efficacy of masks.

That's not the case. For one thing, that study only looked at whether masks protected the wearer, while the main reason for wearing masks is to prevent the wearer infecting other people. And the sample size was too small, which meant that the results weren't statistically significant (which means that you don't know with any sort of statistical certainty whether the results are accurate, and not, as people often assume, that there wasn't any significant difference between mask an non-mask wearers).

As WW says, it was looking at infection rates amongst the wearers and not at rates of infection in the community.


It is incorrect to say that the sample size was too small - it had been calculated to be able to detect a halving of infection rate from 2% (as it was at the time of the study) to 1%. In the event, around 2% were infected in both groups with -0.3% difference between the groups i.e. a higher rate in the mask-wearers; and there is a 95% chance that the difference was from -1.2% to 0.4%. So the accuracy of the trial is not in doubt. There is a 2.5% chance that the masks reduced the infection rate from 2% to 1.6% or less.


The question of a statistically significant difference as against a significant real world difference may be illustrated quite easily. Let us suppose that a morning's operating list consists of 4 patients whose operations take 45 minutes each on average. A study finds that glueing the skin together at the end saves an average of 5 minutes with a 95% confidence interval of 10 seconds i.e. there is a 2.5% chance that less than 4 m 50 s were saved and a 2.5% chance that more than 5 m 10 s were saved. The result is undoubtedly statistically significant; but what are we going to do with the 20 minutes saved? It isn't enough time to do another operation so the difference is not significant in real world terms. (Save that you can have a longer tea break. ? )


ETA: the Danish paper is here

wallywombat:

RB1981:

Indeed the recent Damask-19 research in Denmark found no particular evidence for the efficacy of masks.

That's not the case. For one thing, that study only looked at whether masks protected the wearer, while the main reason for wearing masks is to prevent the wearer infecting other people. And the sample size was too small, which meant that the results weren't statistically significant (which means that you don't know with any sort of statistical certainty whether the results are accurate, and not, as people often assume, that there wasn't any significant difference between mask an non-mask wearers).

As WW says, it was looking at infection rates amongst the wearers and not at rates of infection in the community.


It is incorrect to say that the sample size was too small - it had been calculated to be able to detect a halving of infection rate from 2% (as it was at the time of the study) to 1%. In the event, around 2% were infected in both groups with -0.3% difference between the groups i.e. a higher rate in the mask-wearers; and there is a 95% chance that the difference was from -1.2% to 0.4%. So the accuracy of the trial is not in doubt. There is a 2.5% chance that the masks reduced the infection rate from 2% to 1.6% or less.


The question of a statistically significant difference as against a significant real world difference may be illustrated quite easily. Let us suppose that a morning's operating list consists of 4 patients whose operations take 45 minutes each on average. A study finds that glueing the skin together at the end saves an average of 5 minutes with a 95% confidence interval of 10 seconds i.e. there is a 2.5% chance that less than 4 m 50 s were saved and a 2.5% chance that more than 5 m 10 s were saved. The result is undoubtedly statistically significant; but what are we going to do with the 20 minutes saved? It isn't enough time to do another operation so the difference is not significant in real world terms. (Save that you can have a longer tea break. ? )


ETA: the Danish paper is here