[R] analyzing results from Tuesday's US elections

Matthew McCormack mccorm@ck @end|ng |rom mo|b|o@mgh@h@rv@rd@edu
Tue Nov 17 07:54:46 CET 2020

      No reason to apologize. It's a timely and very interesting topic 
that provides a glimpse into the application of statistics in forensics. 
I had never heard of Benford's Law before and I think it is really 
fascinating. One of those very counter intuitive rules that show up in 
statistics and probability; like the Monty Hall problem. Why in the 
world does Benford's Law work ?  I have been wondering if it could in 
any way be applied to biological data analysis. (Also, I discovered 
Stand-up-maths !).

    Often things are not as easy to figure out as we may first estimate. 
I think you would have to start with how you would envision a fraud to 
be committed and then figure out if there is a statistical analysis that 
could detect it, or develop an anlalysis. For example, if a voting 
machine were weighting votes and giving 8/10ths of a vote to 'yes' and 
10/10ths vote to a 'no'. Is there some statistical analysis that could 
detect this ?  I, Or if someone dumped a couple of thousand fraudulent 
ballots in a vote counting center, is there some statistical analysis 
that could detect this ?  Who knows, maybe a whole new field waiting to 
be explored. A oncee-in-a-while dive into a practical application of 
statistics that has current interest can be fun and enlightening for 
those interested.


On 11/16/20 9:01 PM, Abby Spurdle wrote:
>          External Email - Use Caution
> I've come to the conclusion this whole thing was a waste of time.
> This is after evaluating much of the relevant information.
> The main problem is a large number of red herrings (some in the data,
> some in the context), leading pointless data analysis and pointless
> data collection.
> It's unlikely that sophisticated software, or sophisticated
> statistical modelling tools will make any difference.
> Although pretty plots, and pretty web-graphics are achievable.
> Sorry list, for encouraging this discussion...

More information about the R-help mailing list