An All-Ratio COVID-19 Analysis | Ontario, Canada

-

As a data analyst, I often refer to ratios rather than absolute numbers in analysis. With COVID-19, it shall not be different. Many are the correct yet misleading plots and tables out there. And many are the plainly wrong, deceiving, and manipulated ones. I'll start with a disclaimer that I have absolutely no political reason, motive, or even energy to spend on this from a political standpoint. I am simply looking at Ontario's reported data, and nothing else. The reported data have been sourced at data.ontario.ca/en/dataset/status-of-covid-19-cases-in-ontario. If you are unsure of the terminology, check out the data dictionary provided by Ontario, although I will mention a few of them here. For this analysis, I am looking into the mortality rates between Jun 9 and Oct 9, 2020 (both overall and LTC) as well as the ICU situation in Ontario (both with and without the need for a ventilator).


Overall and LTC (Long Term Care) mortality rates (the number of deaths divided by the total number of cases)

Generically speaking, this is great news. Our mortality rates, both overall and LTC have dropped significantly. It does look like we have been learning about the disease and controlling the deaths that it may bring to a person.



Overall and LTC individual density curves

The density curves, in lay terms, showcase the concentration (probability) of the data being observed. These curves are heavily driven by the high numbers we experienced at the beginning of June and forward (refer to the first plot above). What we want to see from now on, and I think we will given what I explained in the first line plot above, is a heavier concentration of mortality rates on the lower (left) side of the plot.



Overall and LTC overlapping density curves

These are the same as the individual density curves above, however shown in an overlapping perspective. As expected from the data, the mortality rate at the LTC facilities peaked higher than the overall. The good news here is that the curve for LTC has suffered an abrupt stop at those higher rates. That is due to the fact that we have lowered the mortality rate in the long term care homes tremendously since the beginning of June and forward.


Overall and LTC comparative boxplots

These boxplots depict, in this case, the "good outliers" of the mortality rates, specifically, the LTC ones. Those little dots you see at the bottom of the plot on the left side are the really low LTC mortality rates we have experienced post the chaotic situation we saw back in the Spring. This is good, we want to see that entire box (of which 50% of the dataset resides) come down. As far as the overall boxplot goes (the one on the right side), it is good to see the bottom whisker (which represents 25% of the data set) extending further down. Eventually, it will too turn into a bunch of "good outliers".


The ICU situation (the number of ICU-admissions divided by the total number of cases)

In my humble opinion, I believe this is what most health officials are trying to point out. I believe they are afraid of a load & capacity issue, more than an overall death rate. If we look at the ratios for the COVID-related ICU usage, with and without the need for a ventilator, we have a great story to tell up until the end of September. But it is picking up now in October. Once again, and this is just my perception (as many of you have yours, as reporters have theirs, I have mine too), I believe the concerns are more around load & capacity (think of 200 people arriving at the same time to order a latte at the Starbucks drive-thru) than the deaths or complications that the coronavirus is causing to infected people. From a purely operational perspective, this is a legitimate concern.


ICU Situation - individual density curves

As explained in the abovementioned density curves for mortality rates, I'm looking for the probabilistic concentration of rates, in this case for ICU-related admissions with and without the need for a ventilator. It is good to see a higher concentration on the lower ends (left side) however, we have started to see some concentration on the higher end, those bumps you see on the right side of the curves.



ICU Situation - overlapping density curves

Same as per individual density curves, but now in an overlapping fashion.


ICU Situation - comparative boxplots

As opposed to the mortality rates comparative boxplots, what we see here are more outliers beyond the upper whiskers. Those are the ICU-admission rates that we have experienced recently. Those are what I believe to be the main concern health officials have in terms of load & capacity.



by Ramon Perdigao, October 10, 2020, Ontario, Canada. All plots proudly created with RStudio.

R Core Team (2020). R: A language and environment for statistical computing. R
  Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.

Comments

Post a Comment

Popular posts from this blog

The Mathematical Significance of Wisdom Over Time

-
Image
My oldest daughter Maya turned 15 yesterday, Aug 21, 2022. I took some time and went to see her at the end of her special day. Amongst other things, we were discussing wisdom, and how we mature over time. Indeed, a natural process for all of us. I was telling Maya that as she grows older, she will learn things. She will learn things in school, and she will learn things in life. Perhaps more in life than in school. Then we caught ourselves discussing a more numerical type of analysis (Maya has always been amazing with numbers). We realized that, when she was at her year zero (of course, this is a figure of speech) I was 30. She was a baby whom only knew how to cry for food while I knew a little bit about life and things. Now she is 15, she knows a little bit more than what she knew when she was 0, or 5, or 10. I'm 45, I also know a little bit more than what I knew when I was 30 (or 0, or 5, or 10). We decided to graph the data, looking into the "wisdom rate of growth" over...
Read more

Binary Logistic Regression for Raccoon Visits to My Backyard

-
Image
It's that time of the year! No, I don't mean April showers bring May flowers, I mean the time when our fellow wild life animals like to visit our backyards often in search for food. Our raccoon friends have become a bit of an annoyance though since they like to sometimes use our deck as latrine. So, I've been collecting some data and have decided to run a binary logistic regression with the help of SigmaXL statistical package to predict the probability that these masked bandits may (or may not) show up. For that, I've been using a safe and natural repellent (I won't broadcast but, it's coyote urine) to help me in discouraging the presence of the poopy animals around our property. The Data Here's a snapshot of the data I've been collecting. As you can see, the outcome (Y: Raccoon Appearance) expected is a binary (dichotomous) variable. I'm trying to figure out if Mr. Coon and friends will show up or not given certain predictors. I've been trackin...
Read more

Gage R & R Full Example with ProcessMA

-
Image
Gage R&R (Repeatability & Reproducibility) is a Measurement Systems Analysis (MSA) technique that uses ANOVA to assess the accuracy of measurements given variation found in parts and operators measuring parts. We start by looking into the setup of the example. The data contains: Three operators: 1, 2, and 3. Three trials: 1, 2, and 3. Ten parts: 1 through 10. This means 90 responses: each operator measures each part 3 times (3 operators x 10 parts x 3 times = 90 responses). Here's a snapshot of the dataset (first 30 rows): To perform a Gage R&R in ProcessMA, select GageRnR (Crossed) under the option Gage RnR  as shown in the screenshot below. The following fields should be selected accordingly for the study: Study variation: 6 * the standard deviations of each source of variation Hit OK. The first table of results showcases the ANOVA output. Recall from the ANOVA significance tests that we are looking into two or more factors and their interactions; in this example: pa...
Read more