Monday, June 29, 2020

Heading into July

Already June 29, heading to the 4th of July, which in our town always has featured a big (5,000 folk) picnic and fireworks in the Woodlake park, sponsored by Kiwanis and the town, with major local business donors.   Well, this year the 4th of July may come on September 6th, depending on Covid. 

Drat.

It was plenty warm this past week, averaging over 100 degrees.  Saturday was 106, which is hot enough for me.  Not exactly Phoenix or Las Vegas where we've lived on occasion, but still warm.
I'm an inveterate gardener, and my newly planted summer annuals displayed their resentment.

The interactive Dashboard, https://anywhereanytime.io/covid19/, has some metrics for the past two weeks that I found encouraging--446 individual viewers, more than half returned at least once, average Dwell Time was 21 minutes.   And forty+ viewers have written with comments and advice.  THANK YOU ALL for that.

Also, we are five weeks past RE-OPENING America, and we might ask--Howze it goin'?   Answer is NOT WELL.    Here's a plot of the four states getting the most news--AZ, CA, FL and TX.  How do these plots look to you?   Note that on the Maps, ORANGE is bad, Brown is next, Rich Blue is really not good either, while Light Blue and White are GREAT.    Except for rural Northern CA and some of the Texas prairie, the outbreaks have spread across most counties--a new phenomenon



Also, it might be of interest what others think about all this?   The BBC in London posted a couple of stories this weekend, one focused on Texas and the other on the impact for continents of the world.   Hard to assert, as some diehard Trump supporters still do, that this is a hoax designed to drive him out of office--nearly 8 million cases in other countries probably could care less about US politics, but they certainly understand that this is not a hoax, its not going away with hot weather, and it has killed four times the number of any flu strain in memory (the 1918 world event is past our recollection).



Finally, I'd be remiss to not include an update on our local counties--Tulare (my own), Fresno, and Kern, which produce 70% of California's agricultural output.   We've had our own Spikes, it would seem, and people (finally) are wearing masks in our county.   BE SAFE


All in all, let's hope that July exhibits a turn-around for the USA on this Covid19 tragedy--we HAVE to move on, and get some sense of normalcy (was that Warren Harding's only contribution?)

Best regards,
Chuck


Friday, June 26, 2020

Reading Choropleths

You've all seen the geospatial maps on television to describe whee the Covid hotspots are, or how the electoral college might shape up, or which states are sweltering in summer heat.   But you probably didn't call those maps Choropleths.

Yes, that's right.   They have a name!  A choropleth map is a thematic map with shaded or patterned areas in proportion to a statistical variable that represents an aggregate summary of a geographic characteristic within each area, such as population density or per-capita income.   

Edward Tufte self-published his first book in 1975 since no publisher deemed it valuable. The Visual Display of Quantitative Datawas a landmark book for its day.  It had six pages of choropleths, done in grey-scale format, to illustrate geospatial locations for various forms of cancer, by gender.  True, they weren’t explained, and in fact, Tufte got them from the CDC, and in the first five printings, the pictures were misaligned.  But what images! 

I fell in love with the technique, and actually had done a similar thing for the Colorado Air Pollution Commission in 1970.   But it took me a long time to figure out how to utilize computer assistance to build maps.  The result: the http://www.anywhereanytime.io/covid19 InnovaScapes Institute Covid19 Dashboard.

With that Dashboard, we plot four Covid statistics—Confirmed Cases, Deaths, and each of those aggregate numbers as per capita.   We do this for every county in the USA, plus every state, and 225 nations.  The Dashboard has data from the first detection in the USA, January 22, 2020. 

This posting will deal with ‘reading the maps” four weeks after most states “re-opened” across the country, leading to an undesirable but expected surge in cases—May 29 through June 25, 2020.

Let’s begin with Page 2, the states of the USA.   It opens with four maps, as shown below, with a set of bar-graphs at left which show the leading five states for each category, color-coded to match the gradations of the maps.   There are two pictures shown—for the first, notice that the left date slider starts at May 29, so all of the Covid19 data prior to that date has been stripped away.  For the second, the right date slide is at May 28, so this shows everything up until that date. 

Also notice for each set of four maps, there are equivalent bar-graphs to the left of the maps.  These show the top five states in each category.    Note for example in the first set of maps (Figure A), that California and Texas are each a different color in each of the 4 maps, as are Pennsylvania and Florida.  The colors of the bars in the bar-graph correspond to the map colors for any listed state.

                                                                       Figure A

Similarly to Figure A, Figure B shows California and Texas with four different colors, but New York is always Orange, the top color for these choropleths.   Note that all the bar-graph bars are orange, but only a few states on the maps are orange.   
                                                                     Figure B

Next, let's do a side-by-side comparison of the first four months vs. the last four weeks.   Figure C is the total number for Confirmed Cases in each state.   Note that California is Orange in both; no other state is on both lists, so the "Top Five" have indeed experienced some significant shifting.
                                   Figure C -- early and recent CONFIRMED CASES by state

Figure D does the same as Figure C, except that the Cases are normalized by state populations, which gives a much more fair representation of just how likely you might be to become infected.  Here, no state is on both lists, and it is remarkable that the more recent leaders (Arizona, Arkansas, Alabama, South Carolina and Utah) are mostly rural states by comparison with the early leaders (New York, New Jersey, Massachusetts and the District of Columbia).    Note also the clear leader, by a factor of two, is Arizona--a state that proudly lauded its independence from government edict or suggestion.
                   Figure D -- early and recent Confirmed Cases per 100,000 (100K) population

We might do the same for Deaths in each state, and also Deaths per 100K population.   Those are shown in Figures E and F.   New York was the runaway leader in early deaths--who can forget that tragic fact?   The leaders in recent weeks are all at about the same absolute number (Figure E), and in three cases, they are the same states (NY, NJ, MA).   CA and TX replaced IL and PA.  
              Figure E     Deaths for the first four months, and for the most recent four weeks

Note that the maps are very similar for the first four months, between Deaths and Deaths per 100K.   By contrast, the maps for the more recent four weeks reveal a much lower per capita death rate.  How can this be?  Well, with CA and TX in the mix, their high death count is still low in per capita terms
   Figure F    Deaths per capita for the first four months, compared with the most recent four weeks

One more operational note.   If you don't find your state listed in the Top Five, how do you find it?  And in fact, how do you find data for any particular state?   The answer lies in the Selection Table in the upper left hand corner of the main page (Figure 7).   This is an alphabetical listing of every state, which can be scrolled by the scroll bar on the right side, or opened up with the Focus Button in the upper right hand corner (see below).    
Figure 7        Alphabetic list to begin.   Click on any column, and it lists the highest to the lowest,  Click again and it inverts the list to list the lowest first.    Shown here for Confirmed Cases

The effects of scrolling can be seen by reading down the list.  Figure 8 illustrates that California is the clear leader for the most recent four weeks in Confirmed Cases and Deaths, but only 16th in C/100K


Well, a lot of material here, but I hope this gives you a little better idea of how to read choropleths.  And, while we're at it, to navigate the Dashboard with a little more ease.

Spotting the difference

The idea of the Covid19 Dashboard, at http://www.anywhereanytime.io/covid19 with its myriad graphs, tables, and maps, is to be able to discern trends, spot anomalies, and overall give a clear picture of 'what's happening'

So how well does it do that?   We have a test case at the moment--California and Texas are each said to be "exploding" in terms of New Confirmed Cases.   How do we see that on the Dashboard?

Below are two pictures--the first is of the graphs in the lower left-hand corner of Page 1 (the counties maps for USA states), when California and Texas are both selected.   Clearly in the left-most graph, the cases are skyrocketing!   Aha...a solid indication of what the national news is proclaiming
But changing the time-scale is less satisfying, as are the logarithmic plots.

The second is of the county-by-county maps of the two states, shown for the last five months, and then breaking that into the first four months and the past four weeks.   If the past four weeks are 'skyrocketing' you might think that the maps would show that.   And, to be sure, you can find minor differences in each of the maps, but nothing there would cause you to 'catch your breath'

Kinda gives you pause for thought?

It certainly should give evidence of why we insist on 'so many viewpoints'      I could show many examples where a different metric, graph, table, or map illustrates a key difference--but for this instance, there is one outstanding graph to convey the dramatic changes in America at the moment




Thursday, June 11, 2020

UPTICK?

Earlier this week, I posted a listing of the states with sharply rising numbers, along with those with sharply falling numbers.   Now the television stations have picked up on this, and the stock market cratered today, ostensibly because of the stories that Covid is still with us.

Folk have asked for more profiling.   So, here are three graphs to help give some perspective.     This data largely comes from analyses using https://anywhereanytime.io/covid19/

The first graph is of all the states in the nation (dark colors are bad, note that they are heavily in the South and the West, whereas earlier they were in the upper Midwest and Northeast

The second graph shows 7-day moving averages for six smaller states, and then those six plus California and Texas, the two most populous states in the US

The third slide shows seven counties, most of them either Tulare or its closest neighbors.   We clearly are in an UPTICK area, and Tulare itself has just had a resurgence to the highest level it ever had.  The others never saw numbers this high





Monday, June 8, 2020

Impact of re-opening

How soon, people ask, will we know about the impact of "re-opening" states with quarantines or some other version of "lock-down"?   And, of course, there are so many variables that it might indeed be hard to discern, as new organizations and institutions have specific hotspots (e.g. the Fresno prison outbreak this week).

In fact, three 'local' counties exhibit this issue well.   We live in Tulare County, one of eleven roughly contiguous agricultural counties in the area.

Santa Barbara County, with 544 cases total on May 5, had 1,371 cases on May 12.   Of the 827 "new cases", 792 were at the Lompoc prison, 131 of them with guards and staff, the rest were prisoners.

Kings County, with 506 cases on May 23, had already had one food processing plant turn up 170 cases, but by May 31, they had 1,122 cases, and by June 6, 1,492 total cases.   Of the 986 cases in the past two weeks, 878 traced to the prison outbreak.

Fresno County had 1,448 cases on May 23, and two weeks later it had grown to 2,122.   Of the 674 new cases, 614 were at the prison.  

So, for those three counties, 2,284 new cases were specifically at local prisons, but only 203 in the population at large.   By contrast, Tulare County had no prison or nursing home outbreaks of late (they had plenty of nursing home and food processing plant issues earlier), but they had 475 new cases, the bulk of which appear to be 'community spreading'

We're seeing things like this all over America--Iowa, Nebraska, Texas, Oklahoma, South Dakota just to name a few 'unusual' locales.

Here, from our website https://anywhereanytime.io/covid19/ are comparisons of all the states in America, comparing the past two weeks (May 23-Jun 5) vs. May 9-May 22) for Changes.    We only report on States that are UP more than 30%, or DOWN more than 30%.   15 states are up that much or more; 12 states are down that much or more; 23 states and DC are 'more or less' constant.









States Rights

Page 2 of our InnovaScapes Institute Covid19 Dashboard https://anywhereanytime.io/covid19/ is for states and state comparisons, but Page 1 is where the true comparisons are best made in my view

Here, from a sample done two weeks ago, is a comparison of the breadth of Confirmed Cases for ten states--Orange is the 'bad' color, descending to Brown and Rich Blue.   White or Light Blue is GOOD



Note that many of these states exhibit widespread infections while others do not.  WHY?  Is it policy (e.g. the governor of Georgia being 'unaware' of asymptomatic community spreaders) such as early re-opening, or willful disobedient citizens congregating (e.g. "church rights"), or failure to institute 'lock-downs' soon enough?  

Clearly rural counties have, by and large, avoided the levels of confirmed case-rates that more urban centers experienced.   Not surprising in some respects--they travel less, have less interaction with strangers, etc.    It is not true (anecdotally at least) that they abide by social distancing, shelter-at-home, or wearing masks more than their urban counterparts.    But what is happening in Georgia and Alabama and major portions of Colorado obviously suggest that rural areas are not immune.

Sunday, June 7, 2020

California Ag counties and COVID19

Agriculture--feeding us--is big business.   Almost $400 Billion dollars per year in cash crops.  All 50 states produce cash crops, ranging from $30M in Alaska to $50B in California.   Eleven contiguous California counties (of California's 58) account for 75% of California's total agricultural output, and three San Joaquin Valley counties -- Fresno, Tulare, and Kern counties -- are 70% of those eleven.    These three counties alone produce roughly the value of Iowa, the #2 state in America.    Among 30 major crops (more than $100M each), they deliver 90% or more of products like almonds, pistachios, pomegranates, and nectarines for America.  Surprisingly, Tulare County (the smallest of the three) produces one-third as much milk as the Diary State, Wisconsin.  Who'd have thought that?

And this is not counting wine.   Yes, 'AG Cash Crops' plots the grapes sold externally, but wine is considered a processed product, and hence "not agricultural" by definition.   California produces 89% of America's wine, a $700B market.  

But we're here to consider Covid19, and its potential impact on America's food supply.   We plot herein eleven counties, all primarily rural, none more than a million people, half of them Hispanic, with few towns of size, low per capita income rates, and by urban standards, low education achievement and health services.  

Being somewhat isolated has been an advantage for rural areas during the Covid19 pandemic, punctuated by hotspots in food processing plants as well as nursing homes, prisons, and retailing.
And, of course, the absolute numbers are small.   These eleven counties have  five million residents,  just one-half that of Los Angeles County itself.

Thus, meaningful statistics for these counties therefore must be per capita measures rather than absolute.  But trend lines--7 day moving averages, for example -- and Cases per 100K, or Deaths per 100K, can give us meaningful comparison data.

How about an example or two?  These are cumulative for March, April, May and early June.  They are compiled from our InnovaScapes Institute Covid19 Dashboard https://anywhereanytime.io/covid19/

The first example shows the "new case rate" for each county:



The second example shows the per capita Case Rate and Death Rate for the eleven counties, both in table format and in geospatial map.  Tulare, with five to ten times the death rate of the other ten counties (worth asking WHY on that one!), sticks out like a sore thumb on the geospatial map




"The Valley"

For the past eight years, we've lived in "the Valley" for the first time.   We spent many years in "Silicon Valley" which people here call "a little Valley" compared to this "Big Valley"
What they refer to, of course, is the Central California agricultural belt, the "San Joaquin Valley"

Three counties are the heart of this belt, starting with Fresno County at the north, then Tulare County in the center, and Kern County on the south.   These three counties are 'always' the Big Three in terms of annual agricultural revenue--with between 6 and 8 billion dollars each year per county.   Tulare is our home now, the smallest of the three in terms of area, but often the dollar leader.  It is a county with fewer than a half million people, with one town (Visalia) of 130,000 folk, two towns of about 50,000, and seven total incorporated towns.   Our town (Woodlake) is the smallest at 7,600--with something close to 90% Hispanic, and 57% of adults without a high school diploma.

I met this morning with Jessica Reynoso, who graduates from Stanford this week (our first Stanford alum from the local high school in years), and another local man, Gabriel Benavidez, who is a PhD candidate at the University of South Carolina in epidemiology.    He had seen my InnovaScapes Dashboard   https://anywhereanytime.io/covid19/   and had some interesting and challenging questions.   Like "where do we get our data?" and "do we have hospitalization data" and "who are you anyway, our school hasn't heard of Chuck House or InnovaScapes Institute but we follow the Covid data trail with much time and effort.

Jessica starts work in Washington DC next month, working with a non-profit, non-partisan group on issues of governance and equity for gender equality.   At sixteen, she pioneered an Unwed Mother's Community group that won her acclaim in Sacramento (& probably helped her Stanford application).
She was sponsored for several projects by the Woodlake Kiwanis club. Her parents were disappointed when she wanted to leave home for college; today they are incredibly proud of her accomplishments.

Gabriel 'knew' he wanted to go to college, but couldn't afford it.   His folks said, "you can, if you can earn it."   He parlayed a football scholarship into a Physical Therapy major at Montana State U, but he said he found that work not interesting compared to the metrics for sports kinesthetics.   His coach urged him to consider Public Health if he found statistics that interesting.   Baylor University agreed, and granted him a one-year residency for a Master's degree in Public Health.   Then UCLA, UNC, and U of South Carolina offered him a PhD program in Epidemiology, all with the same stipends.  His folks urged UCLA (close to home), but he said his single apartment in Columbia costs $800 per month and a shared apartment with four others at UCLA would be $1,000.   So--costs won out, but he's still proud of Woodlake and things here.   He did wave his arms though and say, I'm probably the only epidemiologist within twenty miles, if not more.   So, I'll have to live elsewhere no doubt.

Why do I relate this story?   Three reasons:
1. Local kids are resourceful, accomplished, and goal-oriented, but they often don't get the credit they deserve, and I just wanted to acknowledge a couple of them that I've come to know.
2. Local kids, to succeed, too often have to leave the area, which is of course a common story for agricultural communities, but it is also a saga of 'brain drain' for areas that need the help
3. Gabriel had incredibly good questions--the best I've gotten-- about our Dashboard, and what data he'd like to see, and how he gets the data he finds relevant.   And that will be the subject of our next set of posts.


Wednesday, June 3, 2020

A month of tuning the Dashboard

I've often got ahead of myself...thinking that "this" as in "this particular project"--must be ready for PRIME TIME.   Only to find that the real learning begins when the first prototype is released.  Such is the case with our Dashboard, first released four weeks ago, and now in its embryonic trial by real users.    Try it, at https://anywhereanytime.io/covid19/ and see what you think.  Fully interactive, county by county, for any date range you'd like--comparative stats are trivial, you'll find.

Where does the data come from?   We use seven sources, primarily USAFacts, the Johns Hopkins Center for Engineering, the 1point3acre Covid 19 site, the New York Times and the Washington Post
All are virtually the same, but occasionally they are out-of-sync, and we pick 'the best'.   Data itself comes from  County Health Departments, and the problems of this are rife--my own county last week announced that they'll only report on weekdays sans holidays.   So Tuesday is usually a "Big Day" in the press, as they roll Friday, Saturday, Sunday and Monday data together.    Counties in some states are forbidden to share Nursing Home data, others asked to forego sharing Food Processing cases--something about causing 'undue fear in citizens'.  

What were our initial learnings?

First, the choice of colors for the Choropleths--red and blue.   Turns out many observers thought this was a political statement, even though the Red Splotches were in fact mostly Blue Urban centers, and the light Blue to Blue areas were more the rural segments of America.  But go figure--everything these days, whether wearing a face mask to thwart (hopefully) Covid19 spreading is seen as unmanly or Left Wing Nuts, or opening a disco tech is seen as cavorting with the devil.  So we changed RED to ORANGE, and kept the various hues of blue.   Not quite as striking for "hotspots" but much more Politically Neutral.

More importantly, the calculation and 'load' time for 3142 counties for all this data is unconscionably slow, so we pre-load California now as the base page, and working from there can be reasonably time-efficient.   Always load a second state before dismissing the first one, will affirm for the program that it only needs to calculate the 2 requests, and not all 50 states.   Then, discard the first state, and voila, you have the new state replete with counties for about 1/25 the time it would be otherwise

And we now offer three pages--US Counties for each state, US States, and the World in combination with US States.   Interactive, powerful, fascinating, and a good way to spend more time than you might think you have available.  https://anywhereanytime.io/covid19/

I plant to issue some YouTube instructional videos--tips for using the site--in due course.   One is now available.   See https://www.youtube.com/watch?v=pp0VsSIBx4Q