2 ms·

I haven't looked in detail, so I can't give you a definite answer, but I offer you this maxim: Random processes clump. When you have things happening at rando

I haven't looked in detail, so I can't give you a definite answer, but I offer you this maxim: Random processes clump. When you have things happening at random you'll find that there are clumps here and there with gaps scattered. If you look at something and find that it's well spread and reasonably even, then it's not random. There will always be some places that have more than average, more than "expected", and there will be places that seem magically to escape. So maybe there's no reason other than it's simply where you are, and it's mostly random. Maybe there is a reason, but really, there might not be. Edit: The comment[0] by gus_massa[1] is completely plausible. [0] https://news.ycombinator.com/item?id=22609125 https://news.ycombinator.com/item?id=22609125 [1] https://news.ycombinator.com/user?id=gus_massa https://news.ycombinator.com/user?id=gus_massa

gus_massa 7y ago
It's a good explanation for a lot of cases and patterns, and it must be the first explanation to try. For example for cancer, that apear with a random distribution, there are spatial and temporal clusters. People usually try to find an explanation for them, but most the patterns are just noise. Even uniform random produce more clustering than what people expect. But in this case I disagree. When there is an initial part of an epidemic with an exponential growing, the time of arrival of the first infected person is very important. It is growing like x2 each week, so an alternative explanation is that the first case arrived to the city one or two weeks earlier. It can be a mix of the two cases. The x2 grow per week is only an average. @OP: Sometimes is good to run a few simulations to see how this work. Start 1000 or 1000000 towns with 1 infected "person" and each day the total number of infected persons is multiplied by a random number between 0.9 and 1.3. After two months, look the distribution of the outcomes of the simulated towns. Most will be near the mode, but there will be some lucky and some unlucky outliers. Note that this is a very bad epidemiology model, and the numbers are totally made up, but is simple and enough to see the clustering. (I made and Exel with 1000 "towns", and I got most of the towns with 10-11-12 ill people. But I got two towns with only 2 ill people and a vey unlucky town with 81 ill people. YMMV) If the initial patient reach the town a different day, you will see that the towns that started the infection earlier usually have more ill people, but some lucky or unlucky towns are in the wrong order when you sort them by the number of patients. Formatted table Name # Population # Contagious =========== ============ ============ Mantova 49400 32 Castiglione 23700 29 Suzzara 21300 4 Viadana* 20150 64 Porto M. 16000 9 Curtatone 14900 11

gus_massa 7y ago

It's a good explanation for a lot of cases and patterns, and it must be the first explanation to try. For example for cancer, that apear with a random distribution, there are spatial and temporal clusters. People usually try to find an explanation for them, but most the patterns are just noise. Even uniform random produce more clustering than what people expect. But in this case I disagree. When there is an initial part of an epidemic with an exponential growing, the time of arrival of the first infected person is very important. It is growing like x2 each week, so an alternative explanation is that the first case arrived to the city one or two weeks earlier. It can be a mix of the two cases. The x2 grow per week is only an average. @OP: Sometimes is good to run a few simulations to see how this work. Start 1000 or 1000000 towns with 1 infected "person" and each day the total number of infected persons is multiplied by a random number between 0.9 and 1.3. After two months, look the distribution of the outcomes of the simulated towns. Most will be near the mode, but there will be some lucky and some unlucky outliers. Note that this is a very bad epidemiology model, and the numbers are totally made up, but is simple and enough to see the clustering. (I made and Exel with 1000 "towns", and I got most of the towns with 10-11-12 ill people. But I got two towns with only 2 ill people and a vey unlucky town with 81 ill people. YMMV) If the initial patient reach the town a different day, you will see that the towns that started the infection earlier usually have more ill people, but some lucky or unlucky towns are in the wrong order when you sort them by the number of patients. Formatted table Name # Population # Contagious =========== ============ ============ Mantova 49400 32 Castiglione 23700 29 Suzzara 21300 4 Viadana* 20150 64 Porto M. 16000 9 Curtatone 14900 11