Coronavirus (non political)

[mike340]

The results I give are extremely conservative (i.e., probably overstate the number of cases continuing past 14 days.) The numbers for total cases were

April 30th: black: 7,821, white: 4,935, and latino: 3,651.
May 14th: black: 11,328, white: 7,479, and latino: 7,864.

The chance that an unrecovered"old case" is out on the street is small (either dead or sick enough so they're not going anywhere.)

This is as opposed to new "unrecovered" cases, where the symptoms may not be incapacitating. Consider the new cases (as of May 14th vs. April 30th):

black: 3,507, and less than half (44.8%) of the cases that existed on April 30th
white: 2,544, and about half (51.6%) of the cases that existed before April 30th
latino: 4,213, which is about 1.15 times (i.e. 115.4%) of the number of cases before April 30th.

Now consider that the proportion of the state population of each group is
black: ~30%
white: ~50%
latino: ~10%.

So the latino numbers for new cases per population* is 8 times what it is for whites and about 3.5 times what it is for blacks. This (along with what I see in the street and my neighbors) leaves me with the general sense they are not getting the message and don't care if they infect somebody. (We worry about some of our latino friends who, for example, have diabetes; I just try to let them know the numbers and find out if we can help them.)

I should also note that the latino rate of increase is about double that of the white and black on EACH DAY, so it's not a statistical anomaly.

*New cases divided by the population pct.

Originally Posted by CRedskinsRule

Mike, I guess as you are a scientist, I am asking what hypothesis you are stating. If I read it right then these would be how I would state your hypotheses:

#1 PG County has the largest section of the outbreak, and should be avoided.

#2 Latinos in Md have been less willing to follow the stay at home order as it is meant to be followed, and therefore non-Latinos should avoid contact with Latinos to reduce the chance of community spread from that specific group.

#3 The Actual Recovered numbers are significantly higher than the reported based on the number of new cases and assuming a 14 to 21 day infection time.

If those are your hypotheses, I generally agree with 1 and 3. Although I think #1 is skewed by the major outbreaks in nursing homes in PG and Riderwood in Montgomery which is actually right on the border of PG if you could somehow pull out those flash points that happened early on, you might see a different statistical variance. But I don't really think that is possible.

I think you are reading the data from number 2 backwards though. I do believe it is significant that the Latino population is less likely to abide by the stay at home order (I admit this is from my walks in the park and the numbers of groups that I see that would be considered larger that 10 or not practicing social distancing so it is not a scientific number). But if we accept the untested assumption that Latinos are less likely to abide by the order, the expected numbers would be an expected higher infection and mortality rate. We see just the opposite in Md. Using your numbers (black: 5,071 cases; white: 3,531; latino: 4,943) Latino and Black active cases are basically equivalent and given the ethnic makeup of PG county, that is not completely shocking. In fact the low number of deaths in the Latino group (black:708; white: 731; latino: 137) should actually have us looking at what factors they have incorporated communally that may have a better effectiveness then stay at home orders alone.

[CRedskinsRule]

Originally Posted by mike340

The results I give are extremely conservative (i.e., probably overstate the number of cases continuing past 14 days.) The numbers for total cases were

April 30th: black: 7,821, white: 4,935, and latino: 3,651.
May 14th: black: 11,328, white: 7,479, and latino: 7,864.

The chance that an unrecovered"old case" is out on the street is small (either dead or sick enough so they're not going anywhere.)

This is as opposed to new "unrecovered" cases, where the symptoms may not be incapacitating. Consider the new cases (as of May 14th vs. April 30th):

black: 3,507, and less than half (44.8%) of the cases that existed on April 30th
white: 2,544, and about half (51.6%) of the cases that existed before April 30th
latino: 4,213, which is about 1.15 times (i.e. 115.4%) of the number of cases before April 30th.

Now consider that the proportion of the state population of each group is
black: ~30%
white: ~50%
latino: ~10%.

So the latino numbers for new cases per population* is 8 times what it is for whites and about 3.5 times what it is for blacks. This (along with what I see in the street and my neighbors) leaves me with the general sense they are not getting the message and don't care if they infect somebody. (We worry about some of our latino friends who, for example, have diabetes; I just try to let them know the numbers and find out if we can help them.)

I should also note that the latino rate of increase is about double that of the white and black on EACH DAY, so it's not a statistical anomaly.

*New cases divided by the population pct.

There are so many factors that simple statistics may or may not account.
For Example,

• you don't have a way to account for time from infection to time or reporting which alone could likely skew one populations number tremendously. I think you could make an effective argument that whites and blacks are more likely to go for medical response sooner than latino families. This effect is for different socio-economic reasons, but clearly isn't accounted for in simply parsing the numbers like you did.
• You don't account for clusters near possible outbreaks that may be unrelated to race ie nursing center that become a major factor
• You don't take into account what actions in the black or white community may have lead to a specific drop outside of community spread.

for me the bottom line is you are taking a personal observation - latinos seem less concerned about social spread (which I agree based on personal observation) and then using statistics in a way that includes many confirmation biases without any use of "clean" statistical data (anonymized and representative) that could confirm or deny your point.