Statistics may be the hardest subject in the world to master. It is incredibly complex and easy to manipulate or screw up with. Unless you have some compelling reason to need a statistical analysis, you are better of staying away from it completely.
Here is the thing about statistics. Like said above the quality of sample data is critical.
Moreover, there are a number of hypothesis tests that can be performed on the data to measure its quality for a given situation. You cannot run these tests if the data is not published with the results.
It just amazes me even back in 2012: there was a sizeable contingent INCLUDING THE MAN HIMSELF who believed Mitt Romeny was going to win because they didn’t understand how to distinguish a junk poll.
It did give me my favorite night of Fox shadenfraude ever though as Karl Rove did his best Don Ameche impression. https://sd.keepcalms.com/i/...
Intercourse the polls. Whatever they say, and however reliable they may be, we have to act like OHJB is two points behind. Let Kornacki spend his time reviewing product surveys for salad dressing.
Seriously, lots of published papers are full of bad stats: scientists get them wrong, professors get them wrong, doctors really get them wrong. (Although in the case of fraudulent hydroxychloroquine study, others highlighted the problems a lot quicker than Gelman did.)
Hoping the Predator drones will notice.
Preach Data Goon, Preach!
If it doesn't come down on the far right of things, it's probably a hoax. Any stable genius will tell you that.
Statistics may be the hardest subject in the world to master. It is incredibly complex and easy to manipulate or screw up with. Unless you have some compelling reason to need a statistical analysis, you are better of staying away from it completely.
Thanks, that was helpful.
There is bad polling and then there is Rasmussen polling (i.e. really really bad polling).
https://uploads.disquscdn.c...
I have more faith in paint
https://twitter.com/emilymb...
Here is the thing about statistics. Like said above the quality of sample data is critical.
Moreover, there are a number of hypothesis tests that can be performed on the data to measure its quality for a given situation. You cannot run these tests if the data is not published with the results.
Bottom line: no data, forget about it.
It just amazes me even back in 2012: there was a sizeable contingent INCLUDING THE MAN HIMSELF who believed Mitt Romeny was going to win because they didn’t understand how to distinguish a junk poll.
It did give me my favorite night of Fox shadenfraude ever though as Karl Rove did his best Don Ameche impression. https://sd.keepcalms.com/i/...
In a perfect world, survey respondents would be half via cellphone and half through a landline.
Really? Do enough people still use landlines to warrant a 50/50 split?
Intercourse the polls. Whatever they say, and however reliable they may be, we have to act like OHJB is two points behind. Let Kornacki spend his time reviewing product surveys for salad dressing.
Seven out of ten doctors said they preferred camels - but most went back to women.
The required statistics class was the rock upon which many college business majors foundered.
It is more complicated than that.
Essentially the goal is to have sampling that is representative of the population in terms of age and other important demographics.
Even though most people have cell phones whether they are young or old, the proportion of younger voters who only have cell phones is much higher.
Therefore, if the split is 50/50 then it to make sure that younger voters are not over counted.
Hi Goon I'm really glad Becca bought you for us
Thank you, Data Goon. You are humble and lovable!
My rule is: "Don't trust any statistics unless Andrew Gelman says it's OK."
https://statmodeling.stat.c...
Seriously, lots of published papers are full of bad stats: scientists get them wrong, professors get them wrong, doctors really get them wrong. (Although in the case of fraudulent hydroxychloroquine study, others highlighted the problems a lot quicker than Gelman did.)