PoliGAF 2016 |OT| Ask us about our performance with Latinos in Nevada

Status
Not open for further replies.
NeoXChaos said:
Why not? She needs every delegate she can get. She's running ads in Bernie's backyard.
Click to expand...

That's because her ad buy guys are the TV ad equivalent of internet trolls and have more money than god. Ad time here is way more expensive so running ads you don't need to is a waste of money.
 
My coin helping Queen Hillary II of BENGHAZZZIIIIIIII First of her Name run them ads.

I paid for 1/4484747th of a second of an ad for her.

Also, all these kids talking about how their mommy and daddy are their heroes and I was rambling about Cher, Liza and Hillary.
 
Samarecarm said:
Sorry, I'm dumb. What are these percentages out of? Like, it's X% out of Y, but what are X and Y here?
Click to expand...

It's the Cook Political Report estimates for where each candidate should be at right now in delegate count in order win the nomination, given the strength and weaknesses of different candidates in different states.
 
CDX said:
I notice they have Bernie winning Nevada. So if he doesn't win Nevada, he'd be under performing according to their model.

Hillary can keep it close and lose Nevada and still be on track for the nomination according to their model.

Interesting.
Click to expand...

Makes sense when you think about it. She's got a decent national lead overall but the polling is pretty close in Nevada. Obviously doesn't take momentum into account though.
 
Also, if these margins hold for Trump, he could sweep all of South Carolina's delegates. I thought he had to win 50% to win all of the delegates, but that's not the case. 29 go to the winner, and then the winner of every congressional district gets 3 more.
 
Samarecarm said:
Is my perception skewed or are there way more than debates this year? It seems ridiculous that there's one every few days. Haven't we hit a point of diminishing returns?
Click to expand...

There were no Democratic debates at all in 2012, so that's probably why you feel like there are way too many now. It does seem over the top.
 
Amy Chozick ‏@amychozick 43m43 minutes ago
Amy Chozick Retweeted Evan Smith
San Antonio Express News endorses @HillaryClinton as race heads to Texas
Click to expand...


It is not that the problems Vermont Sen. Bernie Sanders outlines aren’t real — former U.S. Sen. and Secretary of State Clinton broadly agrees on many of these problems. And it’s not that many of the nation’s ills couldn’t benefit from dramatic reform. It’s that Sanders’ solutions — a single-payer health system and free college, for instance — have no chance to gain traction in what is still going to be a deeply divided Congress after November.
Click to expand...

http://www.mysanantonio.com/opinion...-is-the-best-choice-for-Democrats-6827141.php
 
stolen from the main thread:

AsjIjiK.png
Click to expand...
 
BoboBrazil said:
Realists agree. It's better to choose the nominee that can win in the general instead of someone telling you they are going to give you the world.
Click to expand...


If Hillary loses to Sanders with all her advantages, she would've been a terrible GE candidate anyway.

And all the most successful presidents I remember ran campaigns where they literally promised to give their supporters the world. They all excelled in outlining a visionary future that voters could get excited about. No nonsense practicality is not the stuff dreams are made of, and presidential campaigns are fueled by hope.

Campaign platforms are about laying out what you would do in a perfect world. Most voters understand this. It's a way of laying out what you're ultimate goals are so voters can get a sense for what you're all about.

Hilary is particularly bad at this BTW, and it's hurting her.
 
CCS said:
That does sound pretty interesting. How are you building that model? (Just in general, as someone who's studied stats in some depth but has no practical experience applying it I'm curious) :)
Click to expand...

Bucket features using text analysis (or by hand if engineers didn't really care about digestible analytic logs) and then run logistic regression against customer feature usage to define a very basic model. At this point, I've found models are better at predicting failure as opposed to success.

Take the most significant negative and positive buckets, along with the most highly used ones and do time series analysis (looking at things like time between feature usage in a workflow and find the averages for people who are reaching "success"), look for actions such as undo to indicate failure, and look for save actions to indicate a likelihood of intentional exploration. These can then be used as weighting factors for a given user session as you create a more refined model.

Other approaches such as k-means, etc. can be used to try and find specific user jobs which in some cases should be modeled separately as they have different intent.
 
DataStream said:
Bucket features using text analysis (or by hand if engineers didn't really care about digestible analytic logs) and then run logistic regression against customer feature usage to define a very basic model. At this point, I've found models are better at predicting failure as opposed to success.

Take the most significant negative and positive buckets, along with the most highly used ones and do time series analysis (looking at things like time between feature usage in a workflow and find the averages for people who are reaching "success"), look for actions such as undo to indicate failure, and look for save actions to indicate a likelihood of intentional exploration. These can then be used as weighting factors for a given user session as you create a more refined model.

Other approaches such as k-means, etc. can be used to try and find specific user jobs which in some cases should be modeled separately as they have different intent.
Click to expand...

Cheers, that's really interesting to read! :)
 
DataStream said:
Bucket features using text analysis (or by hand if engineers didn't really care about digestible analytic logs) and then run logistic regression against customer feature usage to define a very basic model. At this point, I've found models are better at predicting failure as opposed to success.

Take the most significant negative and positive buckets, along with the most highly used ones and do time series analysis (looking at things like time between feature usage in a workflow and find the averages for people who are reaching "success"), look for actions such as undo to indicate failure, and look for save actions to indicate a likelihood of intentional exploration. These can then be used as weighting factors for a given user session as you create a more refined model.

Other approaches such as k-means, etc. can be used to try and find specific user jobs which in some cases should be modeled separately as they have different intent.
Click to expand...

The story of my life, right there. Who needs data that can be easily logged and looked back on at some future point, right?

<screams into pillow>

Though Alteryx is surprisingly at managing to pull data out of handwritten files scanned into PDF, as are Python programs.
 
Cybit said:
The story of my life, right there. Who needs data that can be easily logged and looked back on at some future point, right?

<screams into pillow>

Though Alteryx is surprisingly at managing to pull data out of handwritten files scanned into PDF, as are Python programs.
Click to expand...

If there's one thing I've learned it's that I never want to stop dealing with theoretical data sets and have to actually deal with the data that actual people come up with.
 
CCS said:
If there's one thing I've learned it's that I never want to stop dealing with theoretical data sets and have to actually deal with the data that actual people come up with.
Click to expand...

I'm pretty sure 80% of my salary is based on the fact that I have to use actual data, and I never get to even sniff theoretical data to save my life. For 2015; I spent roughly 80% of my time cleaning data and 20% analyzing it. <sigh>

I'm picking up Tableau, R, Alteryx, and SQL though currently. We're kind of splitting up languages / programs through my group for 2016, so we can quickly have a varied set of skills.
 
Cybit said:
The story of my life, right there. Who needs data that can be easily logged and looked back on at some future point, right?

<screams into pillow>

Though Alteryx is surprisingly at managing to pull data out of handwritten files scanned into PDF, as are Python programs.
Click to expand...

I'm dealing with everything from very deliberate logging, people who just dump class names with arguments, people who thought it was a good idea to log the localized version of the feature, and people who log in hex.

Edit: and yeah easily 60% of my job is data munging. luckily the actual work part of my job is 70% experimentation and 30% generating BI.
 
Cybit said:
I'm pretty sure 80% of my salary is based on the fact that I have to use actual data, and I never get to even sniff theoretical data to save my life. For 2015; I spent roughly 80% of my time cleaning data and 20% analyzing it. <sigh>

I'm picking up Tableau, R, Alteryx, and SQL though currently. We're kind of splitting up languages / programs through my group for 2016, so we can quickly have a varied set of skills.
Click to expand...

Such is life I guess. That sounds like a pretty cool way of doing things though!

DataStream said:
I'm dealing with everything from very deliberate logging, people who just dump class names with arguments, people who thought it was a good idea to log the localized version of the feature, and people who log in hex.
Click to expand...

Some of those people have a place in hell reserved for them.
 
Krowley said:
Campaign platforms are about laying out what you would do in a perfect world. Most voters understand this. It's a way of laying out what you're ultimate goals are so voters can get a sense for what you're all about.
Click to expand...

Sanders isn't doing that though. He is promising that he can actually get all this stuff done due to some magical revolution.
 
DataStream said:
I'm dealing with everything from very deliberate logging, people who just dump class names with arguments, people who thought it was a good idea to log the localized version of the feature, and people who log in hex.
Click to expand...

I would laugh; but I just feel bad. I remember my boss telling me 80% of your job is going through a given data set and cleaning it so it becomes usable more so than anything. One of the mandatory things I built into some of the Obama campaign's models years ago was the idea that your data had to be clean, because I'm freaking lazy and don't want to clean that shit. Also, super complicated models as well as cleaning algorithms just get the "glassy eyed" look from superiors, who then proceed to often ignore it.

It's actually why I do sports analytics stuff for fun on the side - SportsVU / etc data sets are SO MUCH CLEANER HOLY CRAP. Though, to be fair, helping out on an update of DVOA was definitely a fun step, as was working with PFF.

NFL is my favorite sport for analytics, in that it's a much more complicated interconnected system, so much more systems engineering and multivariable regressions to play with. MLB almost feels like it's figured out except for the qualitative clubhouse aspects (which I think you might actually be able to get if you figure out a RPM type measure); and NBA has so many awesome folks already that it's just fun to watch them.

That said; I do need to restart some of the DOTA 2 and Street Fighter projects I have sitting on my computer.

My job is all technically BI - but it's more of "combine your aerospace knowledge and your analytics and tell us what's wrong with this, show the numbers, and then tell us how to fix it from an engineering point of view as well as a mathematical point of view."
 
Frank the Great said:
If Trump fails to get 51% of the delegates and loses a broken convention, he's running third party IMO.
Click to expand...

By law he cannot appear on the general election ballot in Ohio as a third party candidate when he was a Republican candidate in the primaries.

I believe Texas is the same.

Most states have "sore loser" laws of some sort.

So, third party Trump run not going to happen.
 
Frank the Great said:
Sanders isn't doing that though. He is promising that he can actually get all this stuff done due to some magical revolution.
Click to expand...

In a perfect world Sanders will use the bully pulpit of the presidency to inspire the masses, vote the bums out in the mid-terms, and take the first real steps in establishing a socialist utopia.

Political revolution.

edit// The future vision Obama painted was almost as idealistic in its own way. He described a post-partisan utopia were all the things dividing us are melted away, and everybody works together for the common good. It was bollocks, but it was a good way of describing the sort of president he wanted to be.
 
A 74 year old socialist President using the bully pulpit to try to get his way will be a great rallying call for the rise of whatever is worse than the Tea Party.
 
Krowley said:
In a perfect world Sanders will use the bully pulpit of the presidency to inspire the masses, vote the bums out in the mid-terms, and take the first real steps in establishing a socialist utopia.

Political revolution.
Click to expand...

In a perfect world, most of what Sanders is advocating for would've already been in place.
 
Status
Not open for further replies.

Similar threads

Sports betting |OT|
7
10K
Hideous Snake replied
ItalianGAF |OT| - Italia
3
950
Azzurri replied
Food |OT| GAFeteria
2
77
17K
SLoWMoTIoN replied
  • Locked
PoliGAF 2016 |OT16| Unpresidented
395 396 397 398 399
20K
1M
PKrockin replied
  • Locked
PoliGAF 2017 |OT5| The Man In the High Chair
395 396 397 398 399
20K
2M
RDreamer replied
Top Bottom