A data scientist's quest for truth (SHORT FILM)
4:12 min. More about Algorithmic biases embedded in PHAC reports. Life story of a senior data scientist.
Produced by Gillian Kirkland
Data by Government of Canada
[PHAC-0] COVID-19 vaccination in Canada, Public Health Agency of Canada:
[PHAC-1] 'COVID-19 Daily Epidemiology Update', Public Health Agency of Canada: https://health-infobase.canada.ca/covid-19/ (https://health-infobase.canada.ca/covid-19/epidemiological-summary-covid-19-cases.html, prior to June 2022)
Data analysis and visualization by Dmitry Gorodnichy and IVIM Inc:
Observation 1: Use of Algorithmic Bias to embellish results
PHAC reports Percentage of fully vaccinated (later of vaccinated with additional doses) among COVID cases counting cases from 14 December 14, 2020, which is when there were NO fully vaccinated (for many weeks yet) and NO additional doses administered (for many months yet).
This technique, known in Data Science under the term of “Algorithmic Bias”, skews the reported results significantly in favour of fully vaccinated (in reports published before June 2022 - by more than an order of magnitude) and in favour of recently administered doses (in reports published from June 2022 - by more than two orders of magnitude).
This Algorithmic Bias can be removed by computing the delta (difference) in the numbers published in consecutive reports. When it is removed this way, the following is observed.
Observation 2: Very different picture, when reported without algorithmic bias
The population with 1 dose (the results for which are NO LONGER reported by PHAC since June 2022 and which counts for 1.2 million of Canadians) has the smallest relative number of Cases, Cases Deceased and Cases Hospitalized since December 2021.
The largest number of Cases, Cases Deceased and Cases Hospitalized in a week per million is observed in populations with additional doses - in both reports issued before June 2022 and new reports issued since June 2022.
When and how many people who completed primary series (formerly called “fully-vaccinated”) did we have in Canada:
How Cases following Vaccination data are reported:
From June 2022 -
Before June 2022 -
What data actually showed:
From June 2022 -
Before June 2022 -
Why the Governments (not only in Canada, but all over the world) are reporting cases following vaccination percentages using algorithmic biases?..
WHO developed these guidelines (for all countries to report cases following vaccination percentages using algorithmic biases) ?..
Can the use of algorithmic biases be ever justified?..
Should the results that are obtained using algorithmic biases be called misinformation, or disinformation, or true information?..
What will it take to make the Governments, and Government of Canada in particular, to admit that they manipulated the ‘Cases following vaccinations’ data so that to “embellish” the results, when publishing them in their weekly epidemiology updates?..
When the Government of Canada will finally start publishing ‘Cases following vaccinations’ statistics without algorithmic biases - they way they publish all other deaths and vital statistics, i.e. by week, or by month, or by year, instead of counting cases since the ‘red ribbon cutting” event of administering the first Covid dose to the first Canadian on December 14, 2020?... We are in Summer of 2022 now…
And a few more:
Was the Rubicon crossed, and if so when? (i.e. when Governments started creating their own “true information” about Covid and Covid vaccines)?…
What does it all say about the Nature of Things and where we are going?…
Maybe you know the answer to any of these questions?..
Comments/questions: dg @ ivim.ca
Appendix: Excerpt from Grievance Document showing the use of Algorithmic Bias.
The Difference in percentages of fully vaccinated reported in August 2021 (which is when PHAC started publishing ‘Cases following Vaccination’ data and which when Algorithmic Bias was the largest) vs. in January 2022 (five months later, which is when the effect of Algorithmic Bias dissipated over time.
Use of advantage (bias) to reduce the percentage of “fully protected” among cases, as there were ZERO fully vaccinated for several weeks after December 14, 2020, and less than 1% before March 2020, which is when majority of cases happened.