This study has gone a bit viral because apparently puppies are going to save our lives
Sadly, that's pretty unlikely to be true 

https://twitter.com/CNN/status/1181558654044119041 …
-
Show this thread
-
I should stay at the start - dogs are amazing, we don't deserve their love BUT Facts are still importantpic.twitter.com/UTslm6eU9g
1 reply 2 retweets 11 likesShow this thread -
1 reply 0 retweets 2 likesShow this thread
-
The researchers did what's know as a systematic review and meta-analysis, looking at all of the studies on dog ownership and health that they could findpic.twitter.com/d6M7e9Xpzu
1 reply 0 retweets 2 likesShow this thread -
They then pooled the numbers from all of these studies into a meta-analysis, which showed that owning a dog was associated with a 24% decreased risk of death, overallpic.twitter.com/J72bvsJGgQ
1 reply 0 retweets 4 likesShow this thread -
So why am I not going out to corral every dog I can find? Well, there were a few significant issues with the studypic.twitter.com/Jjx2YE3zBX
1 reply 0 retweets 5 likesShow this thread -
First up, we have what's known as statistical heterogeneity Basically, this means that the studies that the researchers found and analyzed together were very different, making the analysis less reliablepic.twitter.com/hJwm6VVoLV
1 reply 0 retweets 5 likesShow this thread -
How different were these studies? Well, a commonly-used measure to assess heterogeneity is the I^2 statistic. A value between 0-50% is considered "low" This study found I^2 between 96-98% (except for one subgroup analysis)pic.twitter.com/h078Q4zngn
2 replies 0 retweets 4 likesShow this thread -
Replying to @GidMK
I was seriously puzzled by so many things, including the pooling, I've seen refusal to pool on much much more homogenous estimates. I guess as long as the results are adorable...
1 reply 0 retweets 1 like
Yes it was very odd. I would've just concluded that there's likely too much heterogeneity to trust the point estimates and gone with a narrative synthesis instead I think
-
-
Replying to @GidMK
There recommendations I've always been given is to see if any small subset of studies can be pooled on some criteria, and narrative synthesis all the studies.
0 replies 0 retweets 1 likeThanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.