![\"\"\/](\"https:\/\/lh4.googleusercontent.com\/ue_W0Ow1yRA3-EMMzKGiwPcAnQTrAD9isDOClH7b5RjYJ5Ey0OvcJS8VFz8bib8RNH0saOEljPD5EX_MQkzqT7B7dfHRnG1vNyZIsKJgX4IJdzcmY3LlFshochIaiSPpll-V33T7\")
<\/figure>\n\n\n\nSo we\u2019ve reached 10,000 cases in the US about a day earlier than our projection estimated.\u00a0 What does that mean?<\/p>\n\n\n\n
There are several things it could mean.\u00a0 I\u2019m going to go through a list of possibilities in roughly the order of increasing probability.
1) The virus has gotten stronger.
This is possible, but highly unlikely.\u00a0 To my mind, it\u2019s sufficiently unlikely that I wouldn\u2019t even mention it except that I want to allay any fears about this.\u00a0 Even if the virus had mutated to a more virulent strain, it would start in one location and would then have to spread outward.\u00a0 We haven\u2019t seen any real indication of this.<\/p>\n\n\n\n
2) It\u2019s just noise.<\/p>\n\n\n\n
This is possible and there is undoubtedly some noise in the data, there always is.\u00a0 \u201cNoise\u201d is just randomness in the data. Even though we all know that the weather gets warmer from February to June, if you record the temperature every day, sometimes there will be a few days in a row that get colder.\u00a0 In fact, you\u2019re almost bound to have some periods like that in any given year. It doesn\u2019t mean anything about the warming of the seasons, it\u2019s just noise. But it would take a particularly large amount of noise to cause the difference we\u2019ve seen in the time to reach 10,000 cases.<\/p>\n\n\n\n
3) Sloppy parameters.<\/p>\n\n\n\n
This almost certainly plays a role.\u00a0 When I projected the numbers back on March 14th, I used a starting point of 2500 cases and a doubling period of 3 days.\u00a0 Those numbers were almost certainly already wrong when I hit the button to post. I chose those numbers because they were round numbers that were pretty close.\u00a0 We did cross 2500 that morning, but the real doubling period was unlikely to be exactly 3 days. For example, if the doubling period was really 2.5 days, then we would have projected 10,000 cases for today instead of tomorrow.\u00a0 I chose 3 because it made the projection easy to communicate and my goal was to make some general points about what to expect rather than to make precise predictions. So my choice to use round parameters probably plays a significant role in 10,000 coming early, but I don\u2019t think it\u2019s the whole story.<\/p>\n\n\n\n
4) Increased testing.<\/p>\n\n\n\n
For the last couple of months Coronavirus tests have been hard to come by. I personally know people who had symptoms but could not get a test. As testing increases, we\u2019ll see a rapid increase in cases just because we\u2019re looking more. We may also see an increase in deaths as we look beyond deaths involving flu-like symptoms to what actually causes those symptoms. There are some indications that we may have missed early cases by thinking they were influenza. To see how increased testing can affect apparent growth rates, let\u2019s look at a chart.<\/p>\n\n\n\n