I went to the link, downloaded the spreadsheet, which I linked for convenience. In order to figure out the number of total deaths in 2020/2021 I would say it is straightforward to sum the elements of the `All deaths in England - 2020/2021` columns. The value we obtain is 699471. Let's now do the same operation with the `England - Five-year average`. I think we can think of the result as "the baseline". We obtain 596617. The deaths in England, during 2020/2021, are above the five years average by 102854, a number that is interestingly in the same ballpark as the number of COVID-19 deaths in England, but let's not think too much of it for now. You will have noticed that there is also a `England - Maximum deaths over five years`. We can sum all of these values to obtain 635724, a number still smaller than deaths in 2020/2021, by 63747 to be exact, but now kinda more similar.Gps wrote: ↑Thu Jun 03, 2021 2:00 pmhttps://www.ons.gov.uk/peoplepopulation ... bruary2021merlyn wrote: ↑Wed Jun 02, 2021 6:10 pmYou're possibly not getting the first graph. The sine wave is the baseline number of deaths per week. The red line is the deaths attributed to flu. So to get the number of flu deaths you must subtract the baseline. The Covid graph is simply the number of Covid deaths which is an underestimate because using techniques like the baseline statisticians will calculate the 'excess deaths' which is more accurate and more like the first graph.Gps wrote:Maybe I reading this too fast , because I am also eating but. 1200 a day times 7 makes 8400.
That even below the numbers of 2018 influenza numbers in the top graph. 13000.
One aspect is that flu peaks for a week. Covid just kept going and the points where the figures go down are where strict lockdowns were introduced.
So you were wrong.
To me these numbers look around the same as the number of people who died in 2018.
Were is that pandemic?
So, in any possible way the deaths in 2020/2021 are higher than the last 5 years aggregated figures. However, with the data presented in this table, I think we can do better. Of course, the number of deaths for each year is a random variable: that is why average and max are presented from the last five years. Can we try to figure out whether the 2020/2021 are anomalous with respect the previous 5 years baseline?
The information in the spreadsheets is somewhat limited: we do not know the statistical distribution. So we will need to pick one. We can start off with a Gaussian model. Gaussian models are a good starting point normally due to being a very common occurrence in nature. Let's start with that.
A Gaussian model is identified by two parameters: a mean and a standard deviation. We already have the mean: 596617. The standard deviation is not reported, but the max is. We can then use the Three-Sigma Rule: the probability a value has to be within three standard deviations from the mean is 99.73%. This means, in essence, that we can pick the three-sigma limit as a good figure for the maximum value. So, with that in mind, we set this equation:
Max - mean = 3 * sigma => sigma = (Max - mean) / 3
I.e. we extrapolated a figure for the standard deviation based on the distance between maximum value and mean. With reference on the five years average figures:
mean = 596617
Max = 635724
=> sigma = 13036
Good, we now have parametrized our Gaussian model. This Gaussian model describes then the death toll in "ordinary times", since it is based on the previous five years. Now we can pickup our figure for 2020/2021, which is 699471, and ask ourselves: is this value anomalous with respect our "ordinary times" model?
Or, in other words, if nothing has changed and everything is going exactly as in the preceding five years (which means that the death toll is described by our Gaussian model above), what is the probability of finding, by mere chance (or statistical fluctuation), a death toll equal or higher than 699471? This probability is quantified by integrating our reference distribution, of mean 596617 and standard deviation 13036, in the interval starting from 699471 up to infinity. Since this value, 699471, is 7.9 standard deviations away from 596617, this probability comes down to 0, essentially (it is a value so small it is not even in tables).
OK, so maybe we are being excessive. Lets' be way more permissive. Let's suppose the standard deviation is simply Max - mean = 39107. In this case the value reported for 2020/2021, 699471, is about 2.63 standard deviations from the mean. The probability of observing a value of death toll equal or larger than this would be in this case 0.43%.
So, pay attention to the conclusion now, as it is subtle: the death toll for 2020/2021, 699471 is, with respect any reasonable Guassian model extrapolated by the historical data of the preceding five years, so improbable that has to be considered anomalous. Or better: we fail to find evidence to support that the 2020/2021 death toll is compatible with the previous historical figures.
This, as I said, according to simple Gaussian models. Of course, the analysis can be made more accurate and sophisticated by using the entirety of the raw data and the actual probability distributions, as well as simulation techniques. However:
- I would expect final results to be somewhat similar, as normally Gaussian models offer at least a first good approximation, and;
- I'd say that there isn't a way to look at the data you propose and conclude that the death toll is the same as that of the previous years: all figures for 2020/2021 are in excess of the historic data, the gap from the average is similar to the value of observed COVID-19 deaths, and attempts at quantifying if they can be regarded as compatible with the previous years fail.