docs-website/source/_posts/week-one.md

In epiforecasts/crowdforecastr: Eliciting crowd forecasts in R shiny

title: Predictions from week one of the UK Crowd Forecasting Challenge date: 2021-05-25

On May 24th 2021 the official UK Crowd Forecasting Challenge began and the first predictions are in!

Predictions of Covid-19 infections

The beginning of summer and the continuous vaccine roll-out will likely help curb infections. However, new variants like the B.1.617 strains (the 'Indian variant') as well as the gradual easing of government restritions, mean there may be a risk of rising case numbers. On average, participants in the UK Crowd Forecasting Challenge predicted cases in the UK to rise slightly over the next four weeks until June 19th 2021.

Here is a visualisation of the combined prediction of weekly case numbers that show the central estimate and 50% and 90% prediction intervals (the 50% and 90% prediction intervals are the ranges of values of which forecasters believe they will cover the true observed value with 50% and 90% probability)

The average crowd forecast is very much in line with the average forecast of all model based predictions submitted to the European Forecast Hub by various research groups, even though the crowd prediction is slightly more pessimistic.

This is the Forecast Hub ensemble prediction:

The crowd prediction shown here is of course an average that doesn't show the full range of different predictions. You can see all individual forecasts here. Most forecasters predicted a rise in reported infections, some believed the number of infections would stay flat, and only a minority predicted infections would fall.

Predictions of deaths from Covid-19

Here is the combined crowd forecast of deaths from Covid-19 over the next 4 weeks:

On average, participants predicted no further rise in deaths resulting from Covid-19, but also no decline. This suggests that confidence in the efficacy of vaccines (how well they work) against severe disease is high - even with new variants of concern spreading.

Getting involved

The UK Crowd Forecasting Challenge has just started and you can still join! You don't need to know anything about modelling to take part. To get started, have a look at this post that gives an overview of all you need to know.

If you have any questions, just send us an email at epiforecasts [at] gmail.com

epiforecasts/crowdforecastr documentation built on June 23, 2021, 10:30 p.m.