[EDIT: enough has been learned since I made this model that it is now deprecated. I am now working from home, indeed 2 weeks after the answer was first composed]
A guesstimate model I made to determine whether I should stay home from work. Most of the innovation is in collecting guesses/gut feelings and then doing the calculations. Based on my guesses, I shouldn’t bother working from home for a few months. [EDIT: for complicated reasons it’s probably more like one month] [EDIT: after further adjustments, partly to parameters and partly to the structure, it’s now like 2 weeks] https://www.getguesstimate.com/models/15212
Meta note: the costs of being isolated are roughly linear in how long you’re isolated, but the costs of being in public are exponential in the regime where your chance of getting infected is small and proportional to the number of people who are infected. As a result, you’d rather self-isolate one week early than one week late. Given the model uncertainty in any modelling attempt like this, this means you probably want to be a bit more paranoid than the model suggests.
Model update: previously, the model wasn’t including the possibility of a chain of length more than 1 of people infecting each other at work, ending with you. This increases the disease burden of attending work by a factor of 2.
Model update: There’s been another confirmed case of community transmission in the Bay Area, so I’ve updated the number of infections I expect there to be.
Guesstimate model update: Number of days you want to wait until working from home is (unsurprisingly) very dependent on epidemic doubling time estimates. Fiddling with the distribution to basically update on reports of 1- or 2-day doubling times in some contexts, the model now says that I want to wait a month before working from home.
Well, I tried to calculate that off of current infection levels, but indeed that isn’t built into the model. I agree that’s a big place to change the model.
A non-obvious flaw in the model: the “number of days until you should work from home” distribution is using in its calculation samples from the “current dollar value per day of disease burden of attending work”, rather than the mean as it should. There’s no easy way to fix this, but this pushes that number lower. Note that this error doesn’t affect the calculation of whether I should stay home right now. [EDIT: but it does turn “wait a few months” into “wait a bit over a month”]
[EDIT: enough has been learned since I made this model that it is now deprecated. I am now working from home, indeed 2 weeks after the answer was first composed]
A guesstimate model I made to determine whether I should stay home from work. Most of the innovation is in collecting guesses/gut feelings and then doing the calculations. Based on my guesses, I shouldn’t bother working from home for a few months. [EDIT: for complicated reasons it’s probably more like one month] [EDIT: after further adjustments, partly to parameters and partly to the structure, it’s now like 2 weeks] https://www.getguesstimate.com/models/15212
Meta note: the costs of being isolated are roughly linear in how long you’re isolated, but the costs of being in public are exponential in the regime where your chance of getting infected is small and proportional to the number of people who are infected. As a result, you’d rather self-isolate one week early than one week late. Given the model uncertainty in any modelling attempt like this, this means you probably want to be a bit more paranoid than the model suggests.
Model update: previously, the model wasn’t including the possibility of a chain of length more than 1 of people infecting each other at work, ending with you. This increases the disease burden of attending work by a factor of 2.
Model update: There’s been another confirmed case of community transmission in the Bay Area, so I’ve updated the number of infections I expect there to be.
Guesstimate model update: Number of days you want to wait until working from home is (unsurprisingly) very dependent on epidemic doubling time estimates. Fiddling with the distribution to basically update on reports of 1- or 2-day doubling times in some contexts, the model now says that I want to wait a month before working from home.
It looks like you’re just assuming how many people will be infected, rather than basing that off current infection levels? Is that correct?
Well, I tried to calculate that off of current infection levels, but indeed that isn’t built into the model. I agree that’s a big place to change the model.
A non-obvious flaw in the model: the “number of days until you should work from home” distribution is using in its calculation samples from the “current dollar value per day of disease burden of attending work”, rather than the mean as it should. There’s no easy way to fix this, but this pushes that number lower. Note that this error doesn’t affect the calculation of whether I should stay home right now. [EDIT: but it does turn “wait a few months” into “wait a bit over a month”]
Note of course that this could be missing obvious factors, if so please let me know.