Validation

Posted on

Validation

validation /valɪˈdeɪʃ(ə)n/ noun: validation; plural noun: validations The action of checking or proving the validity or accuracy of something.

As mentioned elsewhere on the site the graph of risk versus infections and attendees is produced using the method detailed by Georgia Institute of Technology on their COVID-19 Risk Assessment Planner website developed for the citizens of the USA. The purpose of this page is to show that their estimation model has been correctly transposed from the US site to this site. It is not possible to input UK data directly into the US site so in order to validate this site against the US model therefore I choose to use Excel as an intermediary to show that the formula have been fully understood and the model used on this site produces results identical to that on the US site.

This first graph is taken from the US site.

Figure 1: Georgia Inst. of Tech. Risk Assessment Planner Output

 

Looking at the horizontal lines at 400,000, 2 million and 8 million infections it can be seen that the following values of risk are estimated:

Table 1: Risk Values for various Infections and Attendee levels – USA data

For a US population of 330 million – the value used by Georgia Institute of Technology – the intermediary Excel model produces the following values:

Table 2: Risk Values for various Infections and Attendee levels, USA Data – Excel output

i.e. the same risk values in both cases – Table 1 and Table 2.

The same Excel model produces the following values for a dataset using a population of England of 56,286,961 (mid 2019 value from UK GOV data) and recorded cases over a 14 day period of 31,181.

Table 3: Risk Values for various Infections and Attendee levels, UK Data – Excel output

and the following equivalent Excel graph:

Figure 3: Risk Plots for Various Infections and Attendee Levels, UK Data – Excel output

Using the same data and formulae this site will produce the following graph (click on Alt View of Data on home page):

Figure 4: Risk Plot for Various Infections and Attendee Levels, UK Data – website output

And the following results from a mouse over event at the highlighted datapoints:

Table 4: Risk Values for various Infections and Attendee levels, UK Data – website output

i.e. the same risk values can be seen in tables 3 and 4 above

(Note that the website is configured to display “<1%” and “>99%”, rather than the actual values, under those circumstances. Also when I started developing the website I followed the Georgia Institute of Technology and used their ascertainment values of 5 and 10 but as time has gone by and we approach the end of Lockdown 2.0 it is apparent that these biases are too high and I have reduced them to 2 and 5 – see the About page). Interestingly I’ve noticed that the Georgia Institute have changed their bias numbers to from 5 and 10 to 3 and 5 (mid Feb-21).

If you want to play around with the Excel model you are free to download the following Excel spreadsheet and do so. Any comments, email me.

Covid-19 Model in Excel