However, any mathematical model - nor matter how accurate or well designed - is only as good as the data that is being fed into it. Unfortunately, the data you guys are getting is poor for three reasons. First, the nature of this disease is such that it is virtually indistinguishable from regular seasonal flu (at least, in the early stages), which means that there may be more sick people running around the world than what the governments and health authorities are actually telling us. Second, it is almost pointless to go by the number of confirmed cases, as the WHO is dreadfully slow and inconsistent in the way they handle the ‘backlogs’. Third, some health authorities are now suggesting that up to one third of the carriers may not even show symptoms at all - but can still infect others (though whether that is true remains to be seen). To add to the confusion, the Mexican government has (for better or worse) ceased reporting new cases of infection entirely, so your data may become even staler than it already is.

So as a fellow researcher who makes a living out of publishing prediction data, but has to live with the not-so-pleasant consequence of getting it wrong (which happens from time to time, nobody is perfect!), my suggestion is this: perhaps you can run a scenario analysis based on different sets initial conditions, with the base case scenario being the official data provided by health authorities. The worst case might assume some sort of multiplier to the reported cases (say, ten times). I understand that this doesn’t sound as exciting as making statements like ‘our calculation shows xx infections by May 17th’, but it’s virtually impossible to make any kind of single point prediction with any degree of accuracy. At the very least, the scenario simulations will allow us to get a better feel for the degree of perbutation based on different sets of initial conditions.

As I mentioned earlier, I do think your team is on to something, and regardless of some of the nasty comments you’ve been getting on this post (it’s normal, everyone’s a critic!) , this project provides valuable insight for the public who is being kept in the dark - as they often are at times like this. I am not sure how long it takes for you to run your models, but my opinion is that an extra bit of information will go a long way in adding value to your hard work.

Again, thanks for your great work, and I look forward to seeing any sort of scenario analysis results.

My question assumes of course, that a significant percentage, if not a majority of probable cases are actual cases. After all, the mayor of New York has stated that the numbers of cases in that city are already likely to be in the hundreds, as they’ve decided not to send in specimens for many probable cases, as they do not want to waste the resources.

Thanks for your time and for providing this analysis.

We welcome your comments and we want to apologize for not being able to respond on each and every of the often interesting and relevant questions and remarks posed in your comments. We do however want to address some in the following:

Linda Watson: We do indeed look at the data in a number of ways. Containment measures as decided by health authorities will be integrated in futures updates. We actively consider social distancing, but surely you will understand that devising meaningful model parameters is a non-trivial exercise. We are also sure that you’ll understand that we have to careful on what we make public and that we cannot satisfy all requests as our resources are currently stretched thin.

Lina: We did run a very special simulation to consider your particular case and we are happy to report that you will not die.

Karen Y: See http://www.newscientist.com/article/dn17046

Mark Malnati: We are projecting cases. The number of reported cases will generally be lower. The R0 parameter in the worst case scenarios is 1.9, meaning that each infected person will, on average, infect 1.9 other persons.

Linda: Each report is based on simulations run with initial conditions derived from the data available to us. As new information becomes available we update the model and run the simulations again. Furthermore, as new cases are confirmed we have to integrate these in the simulations as well. This is why we have and will post updates on a regular basis. As to your second question: I do not understand what you refer to with “The site link …”.

Matt: We are sorry that our reports do no meet your expectation. This is, however, mainly due to a mismatch of your expectations and what these reports actually have to offer. As stated in our disclaimers (first listed here, now provided on this separate page), the results are based on computational models and have to be considered only as an extra source of information and not as the reality of the epidemic unfolding. In other words: we never claimed that our “snake oil” will cure all your diseases, so we dismiss your bite as uncalled for.

Are you projecting cases, reported cases, both ?
What value(s) are you using for the infectiousness of this flu ?

Thanks