Welcome to DU! The truly grassroots left-of-center political community where regular people, not algorithms, drive the discussions and set the standards. Join the community: Create a free account Support DU (and get rid of ads!): Become a Star Member All Forums Issue Forums Culture Forums Alliance Forums Region Forums Support Forums Help & Search

tomm2thumbs

(13,297 posts)
Wed Mar 9, 2016, 07:45 PM Mar 2016

Apparently '538' Never Learns... Latest




Keep regurgitating the same storyline and hope it eventually sticks. And this after they write an article on why the Michigan outcome was so off.


Maybe they ought to turn off their glorified numbers-blender for a day and reconfigure their calculators before they continue to embarrass their vocation.

30 replies = new reply since forum marked as read
Highlight: NoneDon't highlight anything 5 newestHighlight 5 most recent replies
Apparently '538' Never Learns... Latest (Original Post) tomm2thumbs Mar 2016 OP
Actually I'm beginning to believe these polls just incentivize Bernie and his supporters Arazi Mar 2016 #1
Maybe Clinton pays more for good polls tomm2thumbs Mar 2016 #2
538 has a distinct Clinton Bias RunInCircles Mar 2016 #3
Do you know what polls plus is measuring? Renew Deal Mar 2016 #5
It's polls plus..a free ten points to Clinton, right? Godhumor Mar 2016 #10
lol Renew Deal Mar 2016 #11
1 upset in 30 years doesn't invalidate all polling Renew Deal Mar 2016 #4
math youceyec Mar 2016 #6
Don't be silly, of course it does CorkySt.Clair Mar 2016 #7
No shit? TM99 Mar 2016 #29
They've already said they can't account for Michigan until Godhumor Mar 2016 #8
to me, if you know know what caused the last crash tomm2thumbs Mar 2016 #9
They don't know what caused the last crash Cali_Democrat Mar 2016 #16
then they should put that disclaimer on their numbers tomm2thumbs Mar 2016 #19
They already have a disclaimer. nt Cali_Democrat Mar 2016 #23
an accurate and visible one tomm2thumbs Mar 2016 #25
Here Cali_Democrat Mar 2016 #27
yea, it's right where everyone can see it tomm2thumbs Mar 2016 #28
So they'll STILL only poll people with landlines. Ken Burch Mar 2016 #21
538 doesn't poll. At all Godhumor Mar 2016 #22
So they still won't correct their model. Ken Burch Mar 2016 #24
When it is proven to be systemic, they will change inputs Godhumor Mar 2016 #26
What would you have them do? Stuckinthebush Mar 2016 #12
I think 538 is the public face of predictive analysis Godhumor Mar 2016 #13
So, they should admit that then Stuckinthebush Mar 2016 #14
Which is the approach they're taking, I believe Godhumor Mar 2016 #17
Yep Stuckinthebush Mar 2016 #18
What exactly do you want 538 to do? Cali_Democrat Mar 2016 #15
Are you a statistician? randome Mar 2016 #20
Ask Hillary Clinton's own Robby Mook tomm2thumbs Mar 2016 #30

Arazi

(6,829 posts)
1. Actually I'm beginning to believe these polls just incentivize Bernie and his supporters
Wed Mar 9, 2016, 07:49 PM
Mar 2016

Win win for Bernie

So yeah, I'm actually hoping HRC supporters post a TON of these in the next few weeks
.

Only helps us as far as I can tell

RunInCircles

(122 posts)
3. 538 has a distinct Clinton Bias
Wed Mar 9, 2016, 07:52 PM
Mar 2016

Every time you check the difference between polls only and polls plus they increase the lead Clinton has over Bernie in their estimates.
I can understand that the polls only show what they show. Nate's thumb on the scale is that he is assuming any deviation in polled results always breaks Clinton's way.

Godhumor

(6,437 posts)
10. It's polls plus..a free ten points to Clinton, right?
Wed Mar 9, 2016, 08:20 PM
Mar 2016

I kid, I kid.

But, yes, I wish more people would read the methodology notes on 538's forecasting.

 

TM99

(8,352 posts)
29. No shit?
Wed Mar 9, 2016, 09:23 PM
Mar 2016

Nate Silvers has been doing this for 30 years! Damn.

And yes, it is more than one failure. This is the man who failed miserably in his predictions of the UK election last year, and he was certain Sanders had peaked at least three times before last August.

Yeah, I know, the mantra of math. Except math gets beat by psychological biases.

Godhumor

(6,437 posts)
8. They've already said they can't account for Michigan until
Wed Mar 9, 2016, 08:19 PM
Mar 2016

a) They see if the miss is systemic or an anomaly
b) If it is systemic, can the causation be discovered and accounted for

As a result, they're sticking with their developed methodology for the foreseeable future.

tomm2thumbs

(13,297 posts)
9. to me, if you know know what caused the last crash
Wed Mar 9, 2016, 08:20 PM
Mar 2016

you don't continue to put the next shuttle out there

 

Ken Burch

(50,254 posts)
24. So they still won't correct their model.
Wed Mar 9, 2016, 09:03 PM
Mar 2016

Or adjust their forecasts to account for the fact that young voters and people without landlines are being underpolled?

Godhumor

(6,437 posts)
26. When it is proven to be systemic, they will change inputs
Wed Mar 9, 2016, 09:06 PM
Mar 2016

Analysts, and I am including myself in this, never make drastic changes to a predictive model based off one result when all other results have been on-spot. The more data that comes in for the next round of contests will show whether changes are warranted.

That said, I am guessing the midst egregious pollster, ARG will simply be weighted next to nothing from now on (They were also horrendous in 2008).

Stuckinthebush

(10,816 posts)
12. What would you have them do?
Wed Mar 9, 2016, 08:29 PM
Mar 2016

Their job is to develop probabilities based on the data available. All the data pointed to a double digit Clinton win in Michigan. Clearly, the pollsters had the model wrong in that state. So, that provides insight into the model 538 uses. But, what do you think 538 does to develop the probabilities? They aren't uses hunches. They are using data. For the large part the data has been very accurate. This was a weird outlier. They will adjust the models accordingly.

I'm just not sure I understand the animosity toward 538. What should they do with the data?

Godhumor

(6,437 posts)
13. I think 538 is the public face of predictive analysis
Wed Mar 9, 2016, 08:36 PM
Mar 2016

So they bear the brunt in a garbage in, garbage out scenario.

Stuckinthebush

(10,816 posts)
14. So, they should admit that then
Wed Mar 9, 2016, 08:41 PM
Mar 2016

GIGO

They explain that the data are what they are. If pollsters are unanimous,y wrong then their model will suck. But, that also gives them valuable information about weighting of polls.

Godhumor

(6,437 posts)
17. Which is the approach they're taking, I believe
Wed Mar 9, 2016, 08:51 PM
Mar 2016

Starting with observation of the 15th to see if Michigan was an anomaly.

 

Cali_Democrat

(30,439 posts)
15. What exactly do you want 538 to do?
Wed Mar 9, 2016, 08:49 PM
Mar 2016

Their model is based on the polls available. If all the polls are wrong, then there's really nothing 538 can do.

 

randome

(34,845 posts)
20. Are you a statistician?
Wed Mar 9, 2016, 08:56 PM
Mar 2016

If not, on what do you base your conclusion that the polls are wrong? Because one poll last night was wrong, maybe?
[hr][font color="blue"][center]Aspire to inspire.[/center][/font][hr]

tomm2thumbs

(13,297 posts)
30. Ask Hillary Clinton's own Robby Mook
Thu Mar 10, 2016, 05:10 PM
Mar 2016

"We would all be well advised to treat the polling coming out today and throughout this week with skepticism," he said.

Latest Discussions»Retired Forums»2016 Postmortem»Apparently '538' Never Le...