Statistics And Lies

[diddlydo]

The bar charts that we're being shown every afternoon in the press conference are completely meaningless when it comes to daily deaths going up or down as they only cover "new reported deaths", some of which happened days or even weeks previously. How does the government hope to inspire any confidence in its honesty?

14:17
England records 514 more hospital deaths taking total to 16,786
NHS England has announced 514 more deaths of people who tested positive for Covid-19, bringing the total number of confirmed reported deaths in hospitals in England to 16,786. The full details are here (pdf).

Of the 514 new deaths announced today:

111 occurred on 22 April.
216 occurred on 21 April.
75 occurred on 20 April.
The figures also show 110 of the new deaths recorded took place between 1-19 April, and the remaining two deaths occurred in March, with the earliest new death taking place on 25 March.

NHS England releases updated figures each day showing the dates of every coronavirus-related death in hospitals in England, often including previously uncounted deaths that took place several days or even weeks ago. This is because of the time it takes for deaths to be confirmed as testing positive for Covid-19, for postmortems to be processed, and for data from the tests to be validated.

The figures published today by NHS England show 8 April currently has the highest total for the most hospital deaths occurring on a single day: 831."

[fiction-factory]

What lies? You are confusing presenting the only reliable data we have and lies. The figures are published every day. The graph now shows a line that includes care home deaths, but of course it's always about 10 days behind as the cause of death can take many days to be confirmed.
And Chris Witty often refers to what may be a more telling figure- the excess mortality. That is published too.

Can you explain what the lie is?

[Zacs-Master]

Everyone knows (except you) that there's always a lag.

[fiction-factory]

Go on then. Tell us exactly what data they should be publishing in the daily graphs.

[ToraToraTora]

all the figures are available, CSE maths should be sufficient to extract anything you think will tell a better story for you.

[BilSuth]

I don't have problem about the two levels of stats. Obviously, the finalised, and accurate, figures take longer to get. Meanwhile they trot out the daily hospital-only figures simply because it is a more consistent basis to observe trends. But it is more useful (I would suggest) with the new-infections figure, the graph for which runs about seven days or more ahead of the consequential deaths graph.

[diddlydo]

They're showing the bar graph as an indicator of whether things are getting better or worse with regard to deaths per day. That's a lie/meaningless as the figures aren't the true ones for any particular day.

[ToraToraTora]

even if the trend is 3 days behind it still provides the trend.

[ToraToraTora]

so how would a Labour government have provided up to the minute figures?

[jim360]

NHS England deaths/day are updated to show the date of *actual* death, and there are signs in that data that cases and deaths either have peaked or are peaking.

[bhg481]

You have 5 days in which to register a death, so there might be deaths which occurred on any of the days mentioned in the OP which have not yet been recorded, making any of those figures subject to alteration.

[BilSuth]

I agree with your second contribution, Diddly. Deaths stats have too many spikes and dips. I Have to say that even the infection stats, which I prefer to focus on, suffer from some late reporting at weekends for example, with a catch-up a day or two later. For chart purposes I think three-day or seven-day totals are more reliable as they are less volatile because the effect of spikes is averaged out.

[jim360]

Curious that there's a weird blip in the data on March 31st.

[jno]

the Chinese had a lag too, but they were accused (everywhere from AB to the White House) of lying and covering up when they adjusted their figures. Sauce for the goose, sauce for the gander and all that.

[fiction-factory]

some form of smoothing / moving average would help clarify the trend but may be viewed with suspicion by some conspiracy theorists. I am not sure about 31 match, Jim, but I know one big blip occurred when some trusts didn't provide data until a day later. Things like that are sometimes covered in the commentary at the top of the charts

[diddlydo]

2 of the "new" deaths reported in today's total actually happened in March!

[Zacs-Master]

Maybe the Govt should put you in charge, diddly. You seem to be able to root out the crux of all the problems at the moment. Your medical expertise is severely undervalued in my opinion.

[naomi24]

//2 of the "new" deaths reported in today's total actually happened in March! //

Positive news then, diddly. That must mean there are fewer today.

[jim360]

Data seems to indicate that the "peak" occurred around April 10th, in England at least. That seems to be somewhat comforting. What worries me is that the tail seems to be dragging out.