Comments
Twitter Posts Betray Illness
View Comments: Newest First | Oldest First | Threaded View
Page 1 / 2   >   >>
Re: Interesting
  • 4/17/2014 8:12:56 PM
NO RATINGS

LOL @CandidoNIck You've got something there. I think that interenet service providers promise something 99.8% uptime because there are occasions when there is no service.

Re: Interesting
  • 4/17/2014 4:54:31 PM
NO RATINGS

I'm with you there. 99% just means we are covering ourselves for when it fails. Lysol needs to step its game up.

Re: Interesting
  • 4/17/2014 12:13:07 PM
NO RATINGS

@Candido probably that's why Ivory soap went with 99 and 44/100% pure. There's a Straight Dope piece on "pure what?

Re: Interesting
  • 4/17/2014 7:46:50 AM
NO RATINGS

Yes the false negatives have to be an issue as well.  I'm not sure how Twitter is doing the tracking but unless they are actually diagnosed with the flu and there is some way to verify that I think it's tough to map a real world virus spreading this way.  Given the allergy issues around here that come every spring I would also assume that their method has to be good enough to identify common symptoms and toss those out.  A good dumping of pollen could easily skew this if people either assume they feel horrible because of allergens or don't mention that their symptoms are a result of allergies.

Re: Interesting
  • 4/16/2014 7:56:50 PM
NO RATINGS

It's hard to stomach anything that's practically 100%. It just seems... impossible?

Re: Interesting
  • 4/16/2014 10:44:25 AM
NO RATINGS

I also agree. Self reporting is just not unbiased.

Re: Interesting
  • 4/16/2014 10:19:49 AM
NO RATINGS

@SaneIT, 

Yeah Monday mornings are probably just ignored in this studay all together. 

Re: Interesting
  • 4/16/2014 10:06:55 AM
NO RATINGS

@SaneIT points out the problem of 'false positives': People who are tweeting 'flu' and showing picutures of their flu medicines.  To come to the right conclusions in 99% of those cases, does the algorithm ignore any direct mention of flu?

I was thinking of the 'false negatives': People who do actually have the flu, but (for whatever reasons) intentionally avoid any mention of that fact in their twitter feed.

Big data is powerful and I'm sure the right algorithm could make some decent predictions here. If you told me 80% accuracy, I'd find that high but at least possible.  99% just isn't.

Re: Interesting
  • 4/16/2014 8:35:03 AM
NO RATINGS

I agree, change to a population of people who are looking for a reason to be sick and that % is going to drop quickly.  I know some people who seem to have the "flu" two or three times every year.  My opinion is that they'd like some time off so they self diagnose at the first sniffle.  I also wonder how many Monday morning brown bottle flu posts would have to be thrown out to make tracking accurate.

Re: Interesting
  • 4/16/2014 1:52:45 AM
NO RATINGS

@PC, 

I would have to agree, change the population and the accuracy will drop like a rock. 

Page 1 / 2   >   >>


INFORMATION RESOURCES
ANALYTICS IN ACTION
CARTERTOONS
VIEW ALL +
QUICK POLL
VIEW ALL +