One week until Iowa - What is your prediction?

538 has adopted the jem algo...
reweighting the polls = unskewing to proper templates.

It allowed to the jem algo to nail the last election result.



So, I think we have a decent understanding of just about where Iowa is, though not exact. So, my interpretation is the RCPs average did a much better job than 538s weighted aggregate.

So for some inside stuff on polling, 538 reweights polls to correct for error based on what their team sees as modeling errors on what the voter make up will be, ie if a poll weights for 25 percent under 29 year old voters it will “correct” the weight and recalculate.

RCP just averages the polls and I guess the theory is that with enough polls averaged the model is self correcting.

Whatever.

Obviously, both models overestimated Biden’s support, RCP by about 4 points. Also both models underestimated Buttigieg and Klobuchar by about 4 points. Everyone else 3 points or less - including Sanders who was just about right on.

So my take is the polls read where the progressives are going to vote pretty well but missed where the more centrist dem voters are going.

It will be interesting to see if NH goes the same way. It should be an interesting election considering Iowa is still in doubt. Also, Iowa turn was meh. That could be a fluke based on too many candidates but NH can provide confirmation of it not being a fluke if we see 2008 turnout.

There’s actually a lot to say about turnout, too. We saw a big drop off in low education and increase in high education voters in Iowa. This is looking like a Trump effect.
 
538 has adopted the jem algo...
reweighting the polls = unskewing to proper templates.

It allowed to the jem algo to nail the last election result.

You do realize that you are jem, right?

538 sort of resets the skew and RCP doesn’t manipulate the data at all. RCP is looking stronger right now to me.
 
I always wanted to do something so big... I could talk about myself in the third person... like ricky henderson... but this is not that.


my unskewing process eventually became named the "jem algo".
Others referred to it as that too... if I recall.


2. in the past... 538 did an average of the polls...
and they heavily promoted it.

my critique was that during the early months they were averaging garbage.
so the average of garbage is still garbage...


I pointed out the polls had a tendency to unskew in the end and they would say things like the races got tighter in the last few weeks.

Later Nate ackknowlege the polls unskewed and called that "herding".

3. Now... it sure seems like they realize one of the best ways to do this is to unskew the polls from their sometimes ridiculous over sampling of certain segments... and "reweight" the polls to reasonable templates...

hence... the jem algo.



You do realize that you are jem, right?

538 sort of resets the skew and RCP doesn’t manipulate the data at all. RCP is looking stronger right now to me.
 
Thanks for asking Tony... I know you know this... so perhaps you a tossing a softball?

background...

I had a previous post in which nate silver explained that his simulations showed that because hillary's votes were clumped in states like CA Trump could lose the popular vote and still win.

so here is perhaps the best call of a close race in history...

https://www.elitetrader.com/et/thre...-winning-even-without-the-algo.301548/page-84


#833 Nov 7, 2016


"its way too close to call. The algo was quite satisfied with the templates after the herding. So its looks like Hillary has a slight national lead overall by about 1 to 2 points after unskewing and averaging.

So its too close to call.

I find myself and the algo agreeing with Nate Silver's simulations lately.

Because most the national polls unskewed during the herding last week I now see Nate;s work as making sense. Instead of garbage in garbage out as it was before the herding... his work is now very "tradeable".



If you tell me who wins florida, colorado, and Pennsylvania... that is what I will be watching.[/QUOTE]
 
Thanks for asking Tony... I know you know this... so perhaps you a tossing a softball?

background...

I had a previous post in which nate silver explained that his simulations showed that because hillary's votes were clumped in states like CA Trump could lose the popular vote and still win.

so here is perhaps the best call of a close race in history...

https://www.elitetrader.com/et/thre...-winning-even-without-the-algo.301548/page-84


#833 Nov 7, 2016


"its way too close to call. The algo was quite satisfied with the templates after the herding. So its looks like Hillary has a slight national lead overall by about 1 to 2 points after unskewing and averaging.

So its too close to call.

I find myself and the algo agreeing with Nate Silver's simulations lately.

Because most the national polls unskewed during the herding last week I now see Nate;s work as making sense. Instead of garbage in garbage out as it was before the herding... his work is now very "tradeable".



If you tell me who wins florida, colorado, and Pennsylvania... that is what I will be watching.


So you say you nailed it but your call was to close to call? Got it:confused:
 
https://www.npr.org/2020/02/12/8054...troy-price-resigns-after-caucus-night-debacle

Iowa Democratic Chair Troy Price Resigns After Caucus Night Debacle

The head of the Iowa Democratic party filed his resignation Wednesday, as the organization is still picking up the pieces from last week's caucus debacle.

Troy Price had been head of the state party since 2017, but after his role in overseeing a process widely panned as disorganized and opaque, it became an open question whether he would stay on in his job.

"While it is my desire to stay in this role and see this process through to completion, I do believe it is time for the Iowa Democratic Party to begin looking forward, and my presence in my current role makes that more difficult," Price said, in a letter to the Iowa Democratic Party State Central Committee.

upload_2020-2-15_18-55-54.png
 
Back
Top