Why are Trump's approval numbers up?

Exactly!

But this is what the goal of the impeachment inquisition was all about. It was a witch hunt in search of a crime and to sway the American voter.

 

Yes it was a witch hunt. But in the other side the GOP are a Trump cult determined to defend him at all cost. What I think should have happened is we should have investigated first. The democrats could have investigated Trump for his handling, while the Republicans could investigate Hunter Biden. But the democrats want to take Trump down and the Republicans want to take Biden down and defend Trump.

 

Well, that's your opinion. When you get into the business of predicting events and their outcomes, you'll come to understand how difficult it is and how many people would kill for that record. Especially when the average is ~64% and there are many pollsters with significantly less than 78%.

This isn't as complicated as you're making it out to be. Pollsters conduct a survey; average the data; and the average (the sample statistic) is what is used to represent the 45% or 48% number. Let's say, based on the confidence level, we get an MOE of 3%. We don't say explicitly that the population mean is in the center of the MOE. All we say is that when we analyze the data, the average is 'x' with some variability. The MOE just measures that variability and that variability is very important because we want the sample to be representative of the population.

Imagine if someone were to say, "Trump will receive somewhere between 33% to 59% of the vote in the next election." That is a very high MOE, and not statistically meaningful. An MOE of 43% to 48% is provides a better idea of what we can expect. Is it perfect? No, but

Pollsters are only conducting descriptive statistics. There is not very much room for adjustment. There isn't much to suggest that pollsters are having difficulty finding participants. At minimum, we need a sample size of 250 to have a MOE of +/-2.5% or 3, given a confidence level of 95%. Most pollsters conduct surveys with 2,500+. Demographic assumptions are a big issue, but that is not always something a pollster can control. The only thing they can reasonably do is adjust the sampling methodology.

The polls in 2016, nationally, were accurate. Statewide polls were inaccurate. There wasn't anything wrong with the margin of error. The demographic assumptions were incorrect, as you have previously stated. Again, that is not always something pollsters can control.

 

You're describing the confidence interval, not the MOE. The MOE is the range above and below the point estimator or sample statistic within the confidence interval.

You just quoted me describing what the MOE is, so I'm not sure how you managed to get it wrong.

 

You know you have a great candidate when -9.9 is “actually up”.

 

Ahh no. The average is 77% correct.
64% of the pollsters have a better rating.
Solidly mediocre.
People selected for polls are far from random. At least the good ones.

 

You can easily have the mean number different and still be within the margin of error. So his statement is correct. The 2 polls could easily be different by 1 point and be within the margin of error.

 

No. No, it's not. You just made that up. The 538 average is 63% accuracy TTM. Anyone can calculate the information themselves, since 538 makes all of their data public for everyone and in their GitHub repository.

https://github.com/fivethirtyeight/data/tree/master/pollster-ratings/2018

and this is the source file for the poll ratings data:

https://github.com/fivethirtyeight/data/blob/master/pollster-ratings/2018/pollster-ratings.csv

You know, if you really wanted to be a mathematician, perhaps you should have studied a little harder when you attended college.

 

Every poll has it's own margin of error so what you're saying is meaningless.

 

Yep. Same one.
https://github.com/fivethirtyeight/data/tree/master/pollster-ratings/2018

If you just called random numbers and talked to the person that answered the phone, what kind of skew do you think you might get.
Maybe you should think before you post.

 

I already posted that. Not sure what talent you believe it takes to just copy what I post...

Maybe take your own advice because not only does this sentence not make sense, it doesn't even address the prior post in question.

But I guess you're no longer defending your fake statistics so that's a start.

 

Of Course, ANOTHER way of characterizing those polls might be:

WHY is Trump (given this economy) so far underwater?

Given the Economy, Trump's Numbers should be in Mid-60s.

Why are Trump's "supporters" so happy about numbers in the 40s?

Puzzling.

 

The answer is obvious if one bothers to look at the polls in a historical context.

The President started underwater; most Presidents start popular. It only makes sense to judge Trumps popularity based on this context.

 

How about in the "context" of an (apparently) "Strong Economy"?

 

I got my numbers from the same place you did.
Average races called correctly 77%.
277 polls out of 429 were better or equal to Rasmussen. 65%.​

You keep saying over and over random sample. I'm telling you pollsters do not do a random sample.

 

You're not thinking about this the right way. You can argue that the economy should be higher because of a strong economy; but you can also argue that the economy could be lower with a bad economy.

Also, prior President's have had high approval ratings with terribly performing economies, others have had low approval ratings with great performing economies. Remember, the Job Approval rating isn't measuring actual job performance, it's only measuring the perception of actual performance.

 

True.

And "Polls" are only as good as their "methodology".

 

Dude, there are only 397 rows in this dataset. That means there are only 397 unique polls in this dataset. I don't know where you are getting this 429 number from, but I can only assume you made it up, just like you made up everything else. I don't know why you're so content on being so intellectually dishonest. Aren't you never trumpers supposed to have a greater sense of morality?​

What sort of sampling techniques are they using, if it's not random?

 

Yes, I suppose but they're still the best gauge of sentiment we have. If you want to figure out what different types of people feel about a given topic, you take a poll.

 

You seem to be missing the big picture here. Trump has yet to crack 50% approval even with a fairly healthy economy. That means the larger part of Americans aren’t convinced that Trump is a positive influence, which means the slightest downturn will really impact Trump support among the swing voters that determine elections.

 

Lets assume that half of all races aren't very close and its obvious which party is going to win. So that means that 50% of predictions will have a 100% accuracy rate. Now the other 50% of races will have varying levels of certainty from mostly confident to tossup. Lets say that the accuracy here is 50% even including those races that are mostly confident. So the total accuracy of all the races will be 75%. And that 75% is for a really bad pollster who really shouldn't be listened too. Those pollsters with a 78% accuracy, or a 64% accuracy, or 50%, or 0% just aren't very good. Its just garbage data. Sorry its hard for them, but their data is garbage.

Pollsters really can't change their situation and it is true that their demographic assumptions are often incorrect, and this is really hard to control. The problem is that this 3% MOE is assuming accurate demographic assumptions which is obviously incorrect. Your MOE means nothing if you demographic assumptions is wrong, as 2016 taught us.

 

Obama had similar approval ratings at the same time in his Presidency.

 

Obama's approval (12/2/2011) : 43.2%
Trump's approval (12/2/2019) : 42.8%

Both within 0.4 points of each other at the same time in each of their Presidencies. A statistical tie.

Puzzling.

Obama's approval numbers...
https://www.realclearpolitics.com/epolls/other/president_obama_job_approval-1044.html

Trump's approval numbers...
https://dyn.realclearpolitics.com/epolls/other/president_trump_job_approval-6179.html

 

For Obama to be the worst president in history it sure looks like trump is trying to steal that honor.
Obama was at -6.3 and trump is at -9.9
Obama has two popular vote wins and trump has zero
Obama was decreasing the deficit and trump is increasing it

 

I don't understand your rationale. We only record races as a win, loss or contestable. Pollsters are attempting to simply call the result of these races. The number of races called accurately divided by the number of total race predictions is the accuracy rate.

Sure, some races are easier to call than others; however, this has no baring on the accuracy rate because the one has to take into account the number of races a pollster decides to cover over the course of a year. Rasmussen has covered more than 700 races since 2018, the most behind YouGov. This means that it called 512 races correctly.

If you take Siena College, which has only covered 54 races with a 85% accuracy, can you say that Siena College Polling is better than Rasmussen? If so, why?

Again, forecasting is a very tough business and there are plenty of pollsters with +90% accuracy rates. However, having a near spotless record is easier if you don't have much of one.