Prediction: 2660
Actual: 3051
Accuracy: 87.18%
Prediction: 2632
Actual: 2,122
Accuracy: 80.62%
Prediction: 8400
Actual: 6293
Accuracy: 74.92%
Prediction: 4900
Actual: 4023
Accuracy: 82.10%
Sweeeet Caroline…..BAH BAH BAH!
I’ve had Boston (Bahston) in my tracker for a little while now, but never seem to have the time to actually look at anything….and I still haven’t! But I had plenty of time to create a forecaster that looks three weeks into the future.
So I’ll just tell you what my formulas are telling me.
Dave, you suck… Thanks formulas......
Just like everywhere else, Boston had relatively “normal” / “Low” weekly numbers prior to Halloween. Even 4th of July was nothing of serious note. The cases rose by 30% (68 cases) the following week and hit a peak of 543 cases in the week of August 17th. Cases stayed in that general area until the week of October 12th when they jumped to 821 cases. Then 1188 cases the following week. Then the week of Halloween saw 1460 cases and the following weeks saw: 1646, 2115, 2078 and 1966.
The week of 11/9 was 2,115 cases. That’s FOUR TIMES the amount of cases in the second week of October which had 575 cases.
Since then the case count has barely dropped below 2000 a week. (1966 cases the week of 11/23) The biggest spike was the first week of December with 3,180 cases, 193% of the cases the week prior.
Then we get to the week of 12/7, 3,985 cases. Highest yet.
So what do our formulas tell us?
4 Week: 1.29 Spread Rate = 4907 Cases
3 Week: 1.29 Spread Rate = 4911 Cases
2 Week: 1.44 Spread Rate = 5494 Cases
14 Day Average: 580 Avg= 4057.5 Cases
7 Days Average: 655 Avg = 4586 Cases
Admittedly, I’m cheating. This week is already at 1,011 cases in Suffolk County. That’s more cases on in the first two days of the week than any other recent week. With this in mind, the new 7 Day and 14 Day averages are:
7 Day- 666 | 4660 cases
14 Day- 620 | 4341.5 Cases
4300 Cases would be a 3.5% increase from last week. Last week’s increase was 4.6%. So the 14 Day Average Forecast actually looks close to reality. However, all my other formulas are forecasting a bigger increase. And my future model has the week of 12/21 increasing past 6000 or 7000 cases. I highly doubt that will actually be a recorded number seeing all the holidays delaying tests. Most likely the weeks after Christmas and New Year’s will see the brunt of all those missed tests. Especially with all the holiday travel coming up.
I’d say the high case count at the beginning of the week’s means my formulas are probably close to right. However, travel and holidays may delay the actual case recording.
Prediction: 4900 Cases
AFTER NOTES:
Throwing all my Actual and Forecasted Spread rates into a graph, we can see that the Actual Spread Rate is steadily rising.
We also see that my weekly forecasts have stayed within the range of actuals. None of them have spiked outside of the “norm.”
And here is a graph where I lined up all the weekly forecasts. So The 4 Week Average is set back 4 weeks. The 3 Week set back 3 weeks and the 2 Week Average is set back 2 weeks.
Well last week’s predictions weren’t what I expected. My new accuracy rating for all predictions is now at 75.08%, down 11 points. Sheessssh…..See my previous post for all the prediction/actuals information for the week of 12/7.
Please note that I had a Data Quality issue with Rhode Island. It’s explained in the previous post.
I am now tracking FOUR different locations. I feel like I’m playing multiple spots at a blackjack table, except I suck at counting cards. Regardless, this is a great lesson in Data Analytics.
The week of 12/7 saw a case increase of 2,232. 68 cases or 0.03% more than the increase of 11/30. Making the weeks Spread Rate 1.03. The Four Week Average Spread Rate (4WASR) was 1.25.
The 4WASR has now been trending downward over the last 3 weeks.
11/23- 1.33
11/30- 1.27
12/7- 1.25
Previous to Halloween the 4WASR was sporadic and no week seemed to react to the previous. Then we saw it rise consistently from 0.97 to its peak of 1.33 from 10/26 to 11/23. Each week was higher than the previous. My prediction was that we would see a similarly consistent rise after Thanksgiving, adding to a growing wave of new cases; however that was not the case.
Curiously enough, the Two Week Average Spread Rate (2WASR) was the closest to the actual case count with a prediction of a 1.11 Spread Rate and 2410 cases (92.61% accuracy). The 2WASR is also on the decline; its peak was 1.47 for the week of 11/16.
Dave, does a Spread rate decline mean the virus is going away? Not yet. Any Spread Rate above 1.00 means there are more confirmed cases than the previous week had. So the total count keeps rising higher. Once the Spread Rate is below 1.00 then we will see less confirmed cases than the previous week. And a consistently declining Spread Rate below 1.00 is what we need to confirm the viruses Spread Rate is under control. However, all the data I use comes from other sources and is completely reliant on people getting tested and for the test results to be accurate. Meaning, there’s plenty of room for error. Like a large convention center amount of room for error. See last week for example.
FORMULA FORECASTS
4 Week: 1.13 Spread Rate = 2523 Cases
3 Week: 1.09 Spread Rate = 2424 Cases
2 Week: 1.19 Spread Rate = 2652 Cases
14 Day Average: 290 Avg = 2032 Cases
7 Days Average: 319 Avg = 2232 Cases
4 Week Average County Population % Increase: 0.1181% Increase = 1505 Cases
MY PREDICTION
This week is tough, my confidence is shattered. But the show must go on. So Quick and Dirty…
My 4WASR says we are looking at a spread rate of 1.13, translating to 2523 new cases for 12/14. Given the downward trending spread rate over the last three weeks, I believe this number to be possible. However, that would be a 0.12 drop in the spread rate, a drop we haven’t seen before. So seeing as the 12/7 4WASR was 1.25 and the 12/14 4WASR is 1.13, I’ll average those out for a 1.19 spread rate. That spread rate would translate to 2,660 new cases. That number pretty much falls in line with all current trends.
So here goes nothing. My prediction for 12/14 is 2,660.
Well Rhode Island…you did it. The first week I analyze you and you outsmart me. If you see my previous post, you will see that I had a data quality issue with this state. Regardless, even with the new data, my formulas still overshot the mark by 3k.
Rhode Island’s Covid case count for 12/7 was 8545 with a spread rate of 0.96. Again, more than 300 less than the week before…. Something’s up with you East Coast-ers. But I will say, this slump in numbers comes just as the state instituted a ‘Pause’ to try and keep the spread rate down. So maybe its working??
4 Week Graph
3 Week Graph
2 Week Graph
When I throw my Forecasts and Actual Spread Rates in a Line Graph, I get a visual representation of how well the forecasts have done. So far they haven’t flown wildly out of line. (The graph starts at April 20th because that’s when we had enough data to run weekly forecasts)
I won’t spend much time on this one. Obviously I need to examine both formulas and local politics.
FORMULA FORECASTS
4 Week: 1.13 Spread Rate = 9660 Cases
3 Week: 1.12 Spread Rate = 9531 Cases
2 Week: 1.23 Spread Rate = 10517 Cases
14 Day Average: 1200 Avg= 8398 Cases
7 Days Average: 1221 Avg = 8545 Cases
The 7 Day average declined from 1272 over the past week….or 7 days….
4 Week Average County Population % Increase: 0.59% Increase = 5969 Cases
MY PREDICTION
I’m going to blame bad data for last week’s fumble. However, my formulas still predicted a very high case number even with the new, high quality data. I’m going to keep analyzing the data to try and get a more accurate look at whats going on. However….
The average of the all formulas (except the County Population %) comes to 9330.25
That’s lower than the weekly forecasts, but still 800 more than 12/7s total case count.
So with this in mind, plus the states entry into the “Pause,” I will stick with what my 14 Day Average Forecast says. It predicts a lower case count for 12/14, but if the “Pause” is working, then we should see a continued decline.
My prediction: 8400 cases
12/7 in Passaic County saw 2,592 new cases. A 0.89 spread rate from 11/30s 2,927. My prediction was 3,500. Wah wah…….
Passaic County’s Covid Case trend has been different from the other places I looked at. Their steady upward trend started in the first week of September. The only week that didn’t fall in line was 11/16 with a 0.94 spread rate. (The weeks before and after were 1.83 & 1.27)
Based on this trend and the explosive expansion we saw at Halloween, my formulas and I predicted 12/7 to have a much higher case count than it did. Instead, the spread rate went from 1.05 (11/30) to 0.89, about 335 less cases than the week before.
What is the cause of all this? I don’t know. People not getting tested? Lockdown procedures doing their job? I’d be shocked at the later, the first is more believable. Regardless, I doubt this will continue.
FORMULA FORECASTS
4 Week: 1.04 Spread Rate = 2689 Cases
3 Week: 1.07 Spread Rate = 2772 Cases
2 Week: 0.97 Spread Rate = 2510 Cases
14 Day Average: 371 Avg= Cases 2594
7 Days Average: 370 Avg = Cases 2595
4 Week Average County Population % Increase: 0.43% Increase = 2142 Cases
I find it interesting that all my formulas seem to be honing in on a very narrow range of numbers.
MY PREDICTION
I’ll leave this one up to the formulas. The average of all the above comes out to 2550. However, I find that the 4 Week Average County Population formulas always on the low side. So removing that number…. 2632 is the average of the top five formulas.
Unless we see a sudden spike in Covid cases this week, which is very possible seeing as it takes two weeks for some people to see symptoms, I’d say this number is pretty good. It’s an increase from the previous week but it also stays in the cluster of case numbers we have seen these last few weeks.
2632 would be a spread rate of 1.02. More than 12/7 but in line with previous rises we have seen.
My prediction: 2632 Cases
COMING SOON: BOSTON
So this past week was not at all what I expected. I am now tracking four areas, though I only posted three predictions, and every area had a lower number than expected. But some of my formulas were pretty close to right!
Travis County (Austin, Texas)
Prediction: 3400
Actual: 2,232
Accuracy: 65.65%
Closest Formula: 2,410 – Two week average formula was the closest. Three and Four Weeks formula shot for higher numbers. Potentially signaling a slowdown in spread?
Passaic County (New Jersey)
Prediction: 3,500
Actual: 2,592
Accuracy: 74.06%
Closest Formula: 2,141- Four Week population increase average formula was the closest. All other formulas shot in the 3100-3700 range.
Rhode Island (Yea, the whole state)
Prediction: 12,085
Actual: 8,545
Accuracy: 70.71%
Closest Formula: 10,570- Three week average formula was the closest. Four and Two Week averages were too high. Four Week population increase average formula was too low.
***NOTE: I discovered after my prediction that the Rhode Island Data I used was not completely accurate. Googles data for Rhode Island does not update on the weekends. Meaning the following Monday gets all the weekend tests. However, the historical data I used had weekend testing. Therefore my formulas were thrown off with the sudden change in data quality. I will be using the site I received the historical data from for the weekly updates from now on.
Overall weeks Accuracy: 70.14% ……Ouch……
Rhode Island. The state so small, I decided to try to forecast the whole place instead of just a county. Also, it was really difficult to find historical Covid Case data was wasn’t just a chart. For anyone curious of how I get Covid data from past dates, I look for spread sheets with daily Confirmed Case numbers that line up with the data that Google gives me. Until I can find a site with better information than Google, I use their daily case count updates for the weekly increases.
Damn it. I spent an hour analyzing and typing out what was happening in this state and it didn’t save…. Back to the beginning…
The minute I complied the historical Confirmed Covid Case information into my forecaster, I noticed that the spread rate in Rhode Island is higher than any other place I’ve been looking at. But its the smallest state, so it kind of makes sense. You can’t Social Distance in Rhode Island, you can only go like three feet before you bump into your friends Uncle eating Seafood. When I lived in Rhode Island, I shared a driveway with my neighbor….. and somehow only I owned a shovel…in New England. Man I really got the raw end of that deal….Anyways.
Like everywhere else, Rhode Island saw an exponential upward trend in Covid Cases right around Halloween. For example, take a look at the case increase for the nine weeks prior to Halloween. (Why nine? Because I already typed this thing out once and didn’t hit save. And I still have to make lunch for tomorrow and its now 10PM over here…..I ain’t going further back)
10/19- 259510/12- 151210/5- 13819/28- 9719/21- 5189/14- 7829/7- 5798/31- 6168/24 - 640The weeks prior to 10/5 are all somewhat random and within a cluster. Theres no pattern one way or another. But once we get to 10/5, we start to trend upwards. 10/5 saw 1381 new cases for a total case count of 26,748 which is a 5.44% increase from the previous week. Then 1512 cases, a 5.65% increase, then 2595 cases, a 9.18% increase for a total of 30,855 cases. Thats an increase of 5,488 cases in three weeks where as it previously took eight weeks for the same increase!
Then, Halloween happened…..
10/26- 3031
11/2- 3855
11/9- 6702
11/23- 5928
11/30- 8902
34,117 cases in only 5 weeks.
We are now at the exponential explosion stage of the spreading. If you are paying attention to the numbers, we are in a continuous trend upwards with weekly case increases. The only week that didn’t have more cases than the week before was 11/23. This is due to the holidays, there is a noticeable drop in tests conducted on Thanksgiving compared to the days surrounding it.
4 Week: 1.26 Spread Rate = 11,218 Cases
3 Week: 1.187 Spread Rate = 10,570 Cases
2 Week: 1.19 Spread Rate = 10,621 Cases
14 Day Average: 1089 Avg= 7622 Cases
7 Days Average: 1272 Avg = 8902 Cases Notice how the 7 Day average is higher than the 14 day average. This, again, confirms that there is an upward trend continuing everyday.
Previous Week % Increase: 0.8815% Increase = 8902 Cases
Curious, there are two formulas predicting the same case increase count.
There’s no doubt that the cases will continue to rise at a higher rate. Between the unaccounted Thanksgiving/Black Friday cases (Remember, it can take up to two weeks before you may develop symptoms) and all the Holiday shopping (Malls are packed while restaurants are closed….like that makes sense) I believe the upward trend will continue.
So two formulas predict 8902 cases. That’s EXACTLY how many cases we saw in 11/30. I doubt we will see a stall in the increase.
Then I have three formulas predicting over 10,000 cases. The “4 Week” formula accounts for the spread rate for the past four weeks and averages it out. I think its prediction of 11,218 cases is closer to reality. But I’m still not convinced.
If 12/7 saw the SAME percent increase as 11/30, 4.05%…. Then we are looking at 12,949 cases. Which is frighteningly high. Lets split the difference. 12,084.
Rhode Island Covid Cases for 12/7-12/13: 12,085
I really hope I’m wrong about this one. I feel for all my friends and business owners in Rhode Island.
This week I will be working on two different counties. Passaic County New Jersey is my home.
My prediction for the week of 12/6 was 2200-2300. Yea yea, I know. Prediction a range is kind of a cop out. However, the actual count was 2164. So I wasn’t far off. The prediction was based on the reported spike in travel due to the holidays, and following a trend of Holiday Case Spikes. (4th of July, Halloween, etc)
With 2164 new cases, the Travis County Total Case count is now around 40,205. Nearly 8.01% of its reported population.
With that being said, I have started to run multiple forecasts. So far I have: 4 Week Average, 3 Week Average, 2 Week Average, Previous Week Percent Increase (Or decrease), 14 Day Averages and finally….. 7 Day Averages. Not to mention the, no better reason than its interesting, Temperature Tracking. (High Temp = Low Cases. Low Temp = High Cases)
4 Week: 1.25 Spread Rate = 2712 Cases
3 Week: 1.16 Spread Rate = 2518 Cases
2 Week: 1.11 Spread Rate = 2410 Cases
14 Day Average: 247 Avg= 1729 Cases
7 Days Average: 309 Avg = 2163 Cases
Previous Week % Increase: 0.2517% Increase = 1263.25 Cases
Halloween saw a three week long increase streak. The week of Halloween and prior saw case increases of: 669, 595, 691, 466. Yet the three weeks after were 948 (142%), 1444 (216%) and 1824 (273%).
The week of Thanksgiving was 1609. So following the Halloween increase percentages (Thats 1609 X 216%) we could see an increase of 3473 cases for the week of 12/7.
“But Dave, all the numbers you just threw at us are all over the place!" I know. It’s a lot.
FINAL: I’m gonna go with my gut here and say we see an increase close to 216%. This is based on previous holiday trends and putting into account the exponential spread idea that more people spread to more people (Ground breaking idea. I know)
Travis County Covid Cases for 12/7-12/13: 3400
By popular demand (one person asked) I will be tracking Passaic County as well. I really only just started looking at the data. I found historical Covid Case counts starting from March 24th, 2020. I updated the counts using Googles Covid Case Tracker. There is still some work to be done here. However….
Previous week increases were:
12/6- 2927 (Total Case count is 32635. Or 6.5% of total reported population)
11/30- 2927
11/23- 2784
11/16- 2189
11/9- 2325
11/2- 1269
So we can see theres an increase trend. The interesting thing is that the week of Halloween was only 838 cases. The week prior was 468 and the week prior to that (week starting in 10/12 for those of you keeping score at home) was only 371 cases.
So from Halloween we saw an increase per week of:
11/2: 51%
11/9: 83%
11/16: -6%
11/23: 27%
4 Week: 1.27 Spread Rate = 3730 Cases
3 Week: 1.09 Spread Rate = 3185 Cases
2 Week: 1.16 Spread Rate = 3400 Cases
14 Days Average: 381 Avg = 2667 Cases
7 Days Average: 418 Avg = 2936 Cases
Previous Week % Increase: 0.1099% Increase = 1714.50 Cases
Quick and Dirty….. Without spending too much time on this I predict the 12/7-12/13 case increase to be around 3500. It could even hit 4000. My 4 week Spread Rate is high because of the Holiday increase trend. Based on what we saw after Halloween, I predict that forecast to be close to correct.
Below was my prediction for Travis County Covid Case increase for the week of 11/30-12/6. Unfortunately I was out of town that week and could not write up a full run down.
The weeks actual total was 2164. Again, my Guestimates are not far off.
In the end of October 2020 I decided to start tracking Covid-19 Cases for myself. This was both to keep myself educated about the pandemic and to shake the dust off my Data Analytic skills. I was mostly curious if I could find trends in cases and if it was possible to forecast case numbers with any type of accuracy. Keep in mind, everything below is currently based on Travis County (Austin Texas). Other states and cities will be coming.
Covid Cases- Google has a pretty good Covid-19 Case count interface. You can look up any Country, State, City or major County and find their New Case and Death counts.
Temperature- In order to see any trends related to Temperature changes, I needed to find a database of Historical Temperatures for specific areas. I stumbled upon https://w2.weather.gov The site has a database of past Average Temperatures per day. You can look up these numbers by month and are exported into a table which I copied into Excel and rearranged for my purposes.
I also needed future Average Temperatures in specific areas. For this I used https://www.timeanddate.com/weather It gave me the forecasted High & Lows for the next week. I used these numbers to average out each days temperatures and attempted to predict case counts based on future temperature changes.
This is an early view of the Data Processing Excel Sheet. It is in constant layout change processes.
Date | New Cases (Actual) | Total Case Count | Percent of Counties Population | Change in Population Percent from Previous Week | Predicted Covid Cases for the Week | Predicted Total Count | Prediction Error based on Actual | Predicted Total Population Percent with Covid | Previous Weeks Percent Increase | Daily Average Temperatures | Temperature Changes Week to Week | Temperature Change Rating | Actual Week Case Numbers & Percentages
I am currently using two methods to help forecast Covid-19 Case increases:
Based on previous 4 week Case Increase Percentage Averages
Based on Temperature Changes
Educated Guesses
4 Week Averages- Each week I calculate the percent increase from previous weeks. I then use the percentages from the previous 4 weeks and average them out to give me a very basic idea of what next weeks percentage could be.
The idea behind this approach is that if there is a consistent rise or fall in cases over the past few weeks, then this formula will predict a continuation of this trend. However, things like Holidays, drastic changes in weather, or stricter lockdown procedures will disrupt any trend.
This approach is mostly used as my base projections and are not considered to be accurate.
Temperate Changes- Fairly quickly I was able to identify a consistent pattern between Covid Case numbers and weekly average Temperature changes. If the temperatures went up, Covid Case increases went down (Still increased, but the new case count for the week in question was lower than the previous week) If temperatures went down, then the Covid Case increases went up.
My explanation for this is that higher temperatures invite people to go outside where they might be more distanced and the spread of the virus is less likely to happen. Lower temperatures cause people to stay inside and closer to eachother….yuck!
While this approach can’t predict an actual number, it can help me make informed decisions.
Educated Guesses- This is how make my predictions, by using all the above information, historical data and world events. While the data I have gathered can give me a picture of the past, my formulas attempt to give me an idea of the future. But they aren’t full proof.
For example: The formula predictions for the week after Halloween showed only a 567 case increase. However, having seen a 172% case increase the week after July 4th from the base predictions, and knowing that both Holidays typically involve many people gathering together… I knew the base prediction of 567 was low. So I multiplied it by 172% and formally predicited (On Facebook) that the November 2nd to 8th Case increase would be 975.24.
The actual case count ended up being 948. My educated guess was off by 27, a 2.84% error. Thats pretty good if I do say so myself.
However, my prediction of the week to come was very off base. I predicted 800 cases and it ended up being over 1400. My error was not taking into effect two things:
Undiagnosed Halloween cases that were recorded in the following week
The exponential affects of a 172% increase over time. If 1 person usually gets Covid and now 1.72 (lets just say 2)….. and now 2 people got it, then they spread the disease to 2 more people.
So now I am working on a way to track exponential growth and exponential predictions.
Lets see how this goes.