All Lavender room comments aside, it is unfortunate to see another one cancelled/postponed. Interested to see if the pro's that were listed for this race will jump over to IM Tulsa, like Ben Hoffman. Looks like Patrick Lange and Laura Phillips were also on the list for Texas. With it being the North American Championship is it limited to those regional athletes only? As of the most recent list, there were 10 female and 15 male pro's listed for Tulsa. Does anyone know the limit for Pro entry? Tulsa is sold out and has a wait list for age groupers but I know a few that signed up for the waitlist and were contacted pretty quickly about a slot being available and may be worth a shot.

Probably my last mega visualization post... but you made me super curious, so I had to follow-up. I created an expanded version of my simple Obesity scatter chart. This one has a composite score of the 5 higher volume underlying conditions (normalized) that I can flip on and off easily to see how that affects the positioning.

Surprisingly, Age did not have a huge impact. But I have read multiple times that age by itself is not strongly correlated with morbidity; age is just a proxy for one or more of the others. So, this is what it looks like if I flip on the big four: Obesity, Diabetes, Smoking, and Hypertension. Most of the states outside the oval have some obvious external factors, such as the NE states that got it early before we knew how to treat it, Alaska that is remote and low population, Hawaii that closed its boarders, etc. The crazy outlier is West Virginia. That state is by far the unhealthiest in the nation. It ranks #1 in the country for all four of these Covid comorbidities, yet it is well below its pace in the per-capita death position. If anyone wants to praise government leadership, their governor may deserve it.

This is what I used to call "dirty analysis" when I was working with data science teams. But it often lead to formal studies and real discovery. I may reach out to my former Public Health contacts to try to find out if anyone is looking at this that way. If you start reading stories or studies saying that underlying conditions may play a much greater role in determine Covid outcomes than any of our interventions, remember that you saw it here first.

awesome work, kudos.

Thanks for doing this. It is nice to visualize. Can you post the same graph without the red ellipse that you've placed there. I am almost seeing vertical rectangle that goes from 1.25 to 2.25 on the X axis and from 600 to 2500 on the Y axis. Logically we would think that multiple underlying conditions factor in and things should follow the ellipse that you drew. I think in reality there are multiple versions of this where you put down a X and Z axis (example would be age on X, obesity on Z, or age on X, smoking on Z, or obesity on X, smoking on Z). Collapsing them into a co morbidity score you kind of lose the granularity of which factor results in what in what region (I have an entire company that lives on real time data driven energy network optimization, so that's what my teams do, just nothing to do with Covid19, but I am enjoying the graphs you are putting up)

That's not really correct. They accounted for any measures that led to increased "staying at home" metrics.
Ie - gyms closed, people staying home. Schools closed - same. Etc, etc.

---

Next races on the schedule: none at the moment

oooh I like this (I work as a biz analyst, former financial analyst).

I know age didn't have a huge impact on your data set, but what happens if you do put in age for just NE states? Vermont (my state) did very well. It's small and easier to manage. People behaved a bit better than the average. But most of our deaths were elderly. I know CT, MA and NY got covid earlier but would be interesting to see this segment of the country by age. I feel like a manager asking for more :-)

---

Death is easy....peaceful. Life is harder.

Is it your position that there is zero correlation between public health care measures and Covid results?

I am not sure if I know the answer to that. But I do want to understand whether the measures that are considered to be accepted are optimal. This will feed into my decisions - for example, whether to support the recall effort in CA, etc etc etc.

---

Next races on the schedule: none at the moment

Interesting and informative graph. Thanks! My guesses (as a lifelong resident of neighboring Ohio) re: why the WV death numbers are lower than expected, other than leadership of the governor which I haven't read much about: 1.Rural state with residents that don't travel much and a low % of minorities. 2. Potential under reporting of deaths due to C19. Similar reasons for neighboring Kentucky although it has two cities that are much larger than the largest city in WV and I know that their governor was more of a "closer" than several of the governors in neighboring states (TN, IN).

Haha, this is what happened inside my company...I was getting totally frustrated that we had all this raw data and there was no easy way to visualize it from multiple angles and the data lying all over the place on hard drives, google drives, azure drives etc etc....now we have gone from raw data all over the place and sucked it into a universal data base all pre processed by geographic locations all around the world and we have a cloud based system where any of us or our customers can login, access any data fields, any date ranges and produce custom dashboards instantly and see what they want when they want, how they want (but it took me 1.5 years to get from complaining about it to "visualization at the fingertips").

OK back to racing. I am counting on lots of personal challenges, training like a maniac and doing fairly epic training events with like minded friends. Last year on what woudl have been IM Tremblant weekend, I did 10k swimming, 300km bike, with 3800m of vertical and 42.2km of running in a span of 48 hrs. the swim and bike were kind of an ultraman 48 hrs and the run was "Only a marathon" (spread over 4 runs). I will probably do two of those weekends one in July one in Aug

Here you go. And, I bigged it up a little to try to help with clarity. I wish I had the time and resources to do analysis of all the variables to find which ones really do produce the highest correlation. That is exactly the kind of work I have done on other areas in the past. But for now, I must settle for the easy button.

I tinkered a little, but the simple model I created does not work unless all the states are treated the same (I cannot turn on Age for some of them without screwing up everything). But, here is what it looks like with just age switched on. And the age metric is the percentage of people over 65.

Anyone want to guess that odds they would reschedule the same weekend as Waco? Turns out, the 24th is the only weekend that month I absolutely can not race.

After more thought I've realized the major problem with cancellation. I don't like buying t-shirts so I get by wearing my finisher shirts. My 2010 Broadway Bikes Triathlon is starting to look pretty ragged. I need more finisher shirts.

Great work! I'm curious, is this data you collected, or gathered from public sources?

Yes, everything comes from publicly available sites. I try to use origin sources when possible (CDC, NIH, state DoH, etc.). But many of those sites are very hard to download data. So, in the exceptions I use secondary locations that I have verified against source data at some point.

For example, the daily Covid death statistics by state came from Wikipedia. This is because I have not found any other place to find state-level daily data in a structured format that will drop into Excel.

I got the comorbidity numbers from a variety of sources, including CDC, NIH, and national association links. The data scientist caveat is that the years the data were collected are all different, but they are directionally correct. It is not a scientifically valid representation from that PoV (and a few others), but it probably would not be dramatically different if you followed strict scientific rigor.