U.S. Presidential race - Democrat Discussion

[Grizzly Willy]

Past results are no indication of future ones.

Did I say that? I'm asking if he can provide a counter for Mr. Silver's current projection using current data.

I'm not making this argument because I want him to accept the second coming of Lord Bar-Bar the Magnificent, I'm asking because I'm interested in seeing if I'm missing something and shouldn't trust Mr. Silver's analysis.

Hell, if I was biased about this I would be complaining that the analysis doesn't include Gary Johnson.

So 1 minute of searching google and being admonished to "prove him wrong" got me to 50-50.

Imagine if I actually spent some time on it.

Except you missed the part where I stated that predictions can change, and explained why and how.

[Lenonis]

I think the issue here is that (I'm assuming) the foundation of Silver's predictions is a heap of assumptions that things will be status quo between now and the election. That may or may not happen, and it could easily happen that some incident might occur that causes a revision to that prediction that favors Romney or just has Obama doing a clean sweep.

I mean -- let's say for instance Romney's tax records are leaked a week before the election and it shows that he paid no income taxes via creative accounting for the last ten years. Can you imagine the response in the electorate?

Or if Obama decides to invade Iran -- an extremely war weary nation being asked to fight another protracted fight?

So many variables -- which means the prediction is only valid as long as those assumed variables hold true.

[Grizzly Willy]

Exactly. When the variables change, the prediction changes. The prediction is going off of the data we have right now.

[ptwonline]

Exactly. When the variables change, the prediction changes. The prediction is going off of the data we have right now.

Sounds like he needs some sort of Confidence Index based on the odds that a major event happens between the current date and the election date that will strongly tilt the election one way or the other.

So it would be, say, "78% Obama to win, 66% Confidence" or something like that.

[oblivionx]

Exactly. When the variables change, the prediction changes. The prediction is going off of the data we have right now.

Very easy to switch from variables to bias.

Dailykos blogger screams of bias.

[Grizzly Willy]

Can you point to the biases present in his work?

[belfpala]

Very easy to switch from variables to bias.

Dailykos blogger screams of bias.

He kind of graduated from the kos.

Read what I wrote above. If you can find a flaw in his methodology, present it. One flaw was given above, but it's not anything about bias, just the reality of trying to predict things (a flaw present in every prediction).

[oblivionx]

Can you point to the biases present in his work?

Other than having been a blogger for dailykos?

[belfpala]

Other than having been a blogger for dailykos?

Dude.

I know you said you don't click links.

But it was linked earlier. He admits that he tends to vote for Dems, but he hopes that doesn't come through in his model. That kind of transparency is refreshing, really.

[oblivionx]

Dude.

I know you said you don't click links.

But it was linked earlier. He admits that he tends to vote for Dems, but he hopes that doesn't come through in his model. That kind of transparency is refreshing, really.

It's hard to eliminate your prejudices.

[belfpala]

It's hard to eliminate your prejudices.

So point out where the prejudice appears in his methodology.

It's like accusing him of designing PECOTA to make Alex Rodriguez look good, because, shit, he likes Alex Rodriguez.

[Dendrek]

It's hard to eliminate your prejudices.

If ever there was a poster on this forum who proves that point perfectly...

[Lenonis]

Very easy to switch from variables to bias.

Depends -- if he explains the variables and data sources he is using, then it's all transparent and it should be easy to point out where he is biased. Just pointing and saying "bias bias bias" doesn't really mean much.

Although the joke is that science has a liberal bias, the fact is that if you present data and your methodology for producing said data, you should be able to easily spot any bias that would influence the data. (IE -- arbitrarily declaring outliers for instance)

[oblivionx]

At some point you still have to choose what data to value.

For years I've seen bold predictions about X candidate winning be wrong.

[Bobdoletoo]

Exactly. When the variables change, the prediction changes. The prediction is going off of the data we have right now.

I am sorry if I did not read through all of his methodology... but if he is basing his numbers off of polls, and most polls show themselves to over sample democrats, would that not always show obama with a lead?

---------- Post added 2012-09-11 at 02:02 PM ----------

So point out where the prejudice appears in his methodology.

It's like accusing him of designing PECOTA to make Alex Rodriguez look good, because, shit, he likes Alex Rodriguez.

I am a huge baseball fan and huge on numbers, but please don't use an admitted steroid user as your basis

[belfpala]

I am sorry if I did not read through all of his methodology... but if he is basing his numbers off of polls, and most polls show themselves to over sample democrats, would that not always show obama with a lead?

What do you mean by "oversample"? From what I've seen, Dems just tend to be declared more often. So in a random poll you end up with more Dems than Repubs (the rest being undeclared). (and it's like 30% dem, 25% Repub, remainder undeclared).

[Bobdoletoo]

What do you mean by "oversample"? From what I've seen, Dems just tend to be declared more often. So in a random poll you end up with more Dems than Repubs (the rest being undeclared). (and it's like 30% dem, 25% Repub, remainder undeclared).

If you have a 5% more of one party, is that not enough to be over sampled?

I was also looking at nate silver's page. I have some problems with some states. I do not think New Mexico is a 96.6% chance democratic, I think it is lower. I am pretty sure obama wins, but based on recent news, I do not see it as high as he does. I also think North Carolina is closer then 69% republican, I would say more like 55%. Colorado seems high at 73.9%. As does Ohio at 74.9%. I see a lot of these states a lot closer. But I am reading local coverage from the states and not using just polling data.

I also am waiting for the onslaught of ads after debates. Last election was the democrat's push against Bush. This election will be the push against obama. Bush isn't running, but you would not know that considering how many times democrats say his name. They talk more about him then most republicans. I guess you can keep running against a guy who was out of politics for the last 4 years, I do not see him in any of the polling data though. I do know that Romney will be running against obama.

[belfpala]

If you have a 5% more of one party, is that not enough to be over sampled?

Depends. I'm Undeclared (my state doesn't require registration with a party).

(truncated your post because I don't like quoting entire long posts and because I don't have enough information to discuss about your other points).

[Catta]

Depends. I'm Undeclared (my state doesn't require registration with a party).

(truncated your post because I don't like quoting entire long posts and because I don't have enough information to discuss about your other points).

I still cannot fathom why you need to go through all those hoops to vote.

[Kalyyn]

So since this election is already "you're fucked no matter who wins" for me, I think I've identified something that could be a tipping point for me.

It occurs to me that Dr. Afridi is still in Pakistani custody. If either candidate promises to recover Afridi and extract him to the United States, I will vote for that candidate.

Thoughts?