news polls, believe or not

[dross]

Statistics, one of my favorite subjects (and one that I generally had the highest grade in each class). Statistics are a great tool, but are too often used by those that do not even understand the simpliest of topics like the difference between alpha and beta errors.

I took numerous stat classes in the pursuit of math oriented undergraduate and multiple graduate degrees. To this day, I have a 10 volume set regarding statistical science (probably over 100 lb of text).

I clearly see both sides of this issue and I happily note the good and the bad (just like guns).

Some of my favorite statistical comments:

Statistics (properly performed) never lie, but liers use Statistics (only a few of the sources realize their data is biased). A national full color newspaper was noted in several classes on the improper use of statistics for their color charts or graphs (only showing the ends of bar charts, generating a bias in the way the data was interputed by readers that then expressed their views in polls, perhaps even well conducted ones?).

The doctor to his worried transplant heart recipient (when hearing there was a 50% chance of survival); have no fear the guy yesterday died so you should be OK.

Garbage in = garbage out (some here have noted the joy they received in harassing the harassers)

63.4% (or pick your favorite number) of data presented is made up on the spot! (I think Dilbert is the source).

Being prequalified to take the poll (asking prequestions with answers that coorelate with their targeted response) - creating intentional bias.

There is some excellant use of statistics, it is just really hard for an individual to know if they are seeing the results of it. There are very accurate predictions of voting results, but is tallying self-reported voting after the fact (upon leaving the voting booth) a prediction or just the proper use of statistics to report what just happened?

Statistics are not good or bad, but a combination of math (what is thought to be truth) and science (thoery that seems to represent the world we live in) that produces a tool (sometimes used well and sometimes not).

Guns are not good or bad, but a combination of metals and propellants (sorry plastic lovers) that produce a tool (sometimes used well and sometimes not).

In each case:

There are experts
There are novices and beginners
There are illegal acts (knowly and unknowly)
There are the clueless

Guns and statistics should not generate emotion themselves, only the acts of people using them should be judged.

But I can tell good versus bad statistics; it only depends on if I agree with the results!

I'm like 58% sure that 70% of this post is 90% bull****. No, make that 59.6%.

 

[melensdad]

When are those polls taken? Timing is everything. I said a snapshot. If you want to know what I look like now then the picture taken last weekend is a better representation than the one at my senior prom, is it not?

Not sure how many times I have to retype the same basic information but this will be my last time. Research the polls. Understand their methodology. Be skeptical but be open minded enough to know when to believe.

In the case of Rasmussen, they take the same presidential poll EVERY DAY, they then post a 3 day running average, EVERY DAY, that shows a pretty darn good snapshot of the mood of the voters, it also provides a trend line of the direction of the mood. Is last years daily tracking poll useful for today? Clearly not. Is last weeks information, when combined with this weeks information a reasonable predictor of what the mood will be over the next week? Yes.

Now if you then combine data from a reputable source like Rasmussen, with the data from another reputable source like Pew and if that data corresponds with the data of yet another reputable source then its probably pretty fair to say that the data is sound. That is what the Real Clear Politics poll averaging score does. R.C.P. doesn't actually do any polling, it simply takes polling data and compares it with other polling data from similar time periods, and it averages the polls.

If you don't like what the polls say then research the polls. If the polling company is generally considered to be reliable then you are the one who is out of touch.

 

[88GT]

Not sure how many times I have to retype the same basic information but this will be my last time. Research the polls. Understand their methodology. Be skeptical but be open minded enough to know when to believe.

Not sure why you feel the need to retype it in the first place. Is anything I've said contradictory to your position?

In the case of Rasmussen, they take the same presidential poll EVERY DAY, they then post a 3 day running average, EVERY DAY, that shows a pretty darn good snapshot of the mood of the voters, it also provides a trend line of the direction of the mood. Is last years daily tracking poll useful for today? Clearly not. Is last weeks information, when combined with this weeks information a reasonable predictor of what the mood will be over the next week? Yes.

If that were the case then changes in favorable or unfavorable status would never reverse direction. Or more specifically, we would never expect them to if all we have to rely on is previous poll data.

Now if you then combine data from a reputable source like Rasmussen, with the data from another reputable source like Pew and if that data corresponds with the data of yet another reputable source then its probably pretty fair to say that the data is sound. That is what the Real Clear Politics poll averaging score does. R.C.P. doesn't actually do any polling, it simply takes polling data and compares it with other polling data from similar time periods, and it averages the polls.

I think you're confusing accuracy of the data with something else. A poll can't tell you anything but what that group of people was thinking at that particular point in time. Ask the same question of the same people 2 weeks later and you aren't guaranteed the same results. In fact, you probably won't get them. People are affected by numerous factors when they take those polls. It's why the Kenyan's approval rating can increase 4 percentage points seemingly overnight....from one single event....taking place halfway across the world.

It's still nothing more than a representation of a single point in time. It may be very accurate because it was well designed and the interrogators were very good. But it doesn't tell you anything more than what those people thought on that day. It doesn't tell you why they answered yes or no, or strongly agree or somewhat disagree. It doesn't tell you if they were having a bad day because their pet alligator just died and the whole world looked bleak or whether she was getting married tomorrow and saw everything through rose colored glasses that day. It only tells your what she thought on that day. Come back in 2 weeks when she finds out her new hubby cheated on her with the bartender on the honeymoon and I'll bet you get slightly different answers.

Running averages are great, but only for tracking trends. Not predicting the future. A well-designed poll's results may be so accurate in representing an opinion that it appears to predict the results, but it doesn't really. We make the leap of logic. We say, If this poll is reflective of the big picture of Americans, and if the election were held now, we would expect to see these results. That's not a prediction. That's an educated guess. Psychics make predictions.

Besides, trending in poll results is risky. If what you say is true about results in the immediate past "predicting" results in the immediate future, upward or downward trends would never change. Or the fact that they did would be unexplainable.

If you don't like what the polls say then research the polls. If the polling company is generally considered to be reliable then you are the one who is out of touch.

Why do you keep intimating that I dislike what the polls have to say?


This has nothing to do with any given polls particular results. I'd have this criticism of a poll that said 100% of the people think bacon is better than sausage on a pizza. Because I know tomorrow someone who's never had sausage is gonna try it and decide he likes it better.

 

[dross]

. Ask the same question of the same people 2 weeks later and you aren't guaranteed the same results. In fact, you probably won't get them.

No, but if you ask the same questions for YEARS and then look at what happened in those years, you can get some information that can be used more generally than just for a particular point in time.

It's like a weather report. It's not a perfect predictor, but as much as we all like to make fun of the weather report, it's pretty close to right more often than it's not. They use a similar method. They look at the particular conditions and the type of weather those conditions have produced in the past. That's why you get a 75% chance of rain. Of course, a 75% chance of rain also means that it won't rain 25% of the time given those same conditions.

Well done polls tell you lots of things accurately. They just don't tell you what everyone wants to be told, which is who will win a particular election.

 

[88GT]

No, but if you ask the same questions for YEARS and then look at what happened in those years, you can get some information that can be used more generally than just for a particular point in time.

It's like a weather report. It's not a perfect predictor, but as much as we all like to make fun of the weather report, it's pretty close to right more often than it's not. They use a similar method. They look at the particular conditions and the type of weather those conditions have produced in the past. That's why you get a 75% chance of rain. Of course, a 75% chance of rain also means that it won't rain 25% of the time given those same conditions.

Well done polls tell you lots of things accurately. They just don't tell you what everyone wants to be told, which is who will win a particular election.

I didn't say they weren't accurate. All I said was that any given poll was nothing more than a snapshot of the reality at that time. Now I feel like I'm repeating myself. Besides, accuracy is a completely different component of this discussion and has nothing to do with the point I am making.

 

[dross]

I didn't say they weren't accurate. All I said was that any given poll was nothing more than a snapshot of the reality at that time. Now I feel like I'm repeating myself. Besides, accuracy is a completely different component of this discussion and has nothing to do with the point I am making.

It's clear then that I don't understand your point.

A poll is a snapshot of reality at a certain time, yes.

Nothing can predict the future because we don't know what will happen in the future.

If you flip a coin and get heads ten times in a row, the odds it will be heads on the next flip are still fifty fifty. Yet the chance that you'll flip a coin eleven times in a row is very low.

If you take a poll it's possible someone just had a bad day. It's unlikely that you randomly caught all people who had a bad day.

A poll can't predict who will be president. There are examples of the polling being wrong. Yet, there are things the poll can tell you that enable you to know the likelihood of someone being president.

The earlier it is, the less you can tell about the outcome of the election for the simple reason that way too many things will change public opinion before the election, the most important of which is that the "swing voters" or as I like to call the, the morons, don't even make up their minds until October. Yet you can tell certain things, like, "No one with these numbers has ever done such and such."

Like in the NFL. Every year you'll get something like, no team who has gone 0 and X has ever made the postseason since, 19XX. Yet someday a team will do that very thing.

 

[other guy]

In all my years I have lived all across the United States, and I have never been called for a poll for anything, not on policticals, drugs use, guns, nothing. Then you looks at the numbers given for the poll and they say they polled 1,000 people, who in the hel@ would really believe that a thousand people can speak for the entire U.S.!! I've always said if you really want to know what most people think put in question on the election ballet, then you would at least have the say of anyone who bothered to vote. I've always known that anyone that wants to can play with numbers to get them to come out anyway THEY want, so why waste their time and money???

I also have never been polled for a poll. BUT.... Somehow they almost always pick the winner. I don't know how many times the candidate I supported was behind in the polls and I thought "well maybe the polls are wrong " Well, guess what, they always seem to be right. Now I'm talking about polls for candidates for office, which you can tell the results for, one either wins or not. Unlike opinion polls where you will never really know one way or another. Like I say, I have hoped the polls were wrong many times, but it is rare when they are.

 

[88GT]

It's clear then that I don't understand your point.

A poll is a snapshot of reality at a certain time, yes.

Nothing can predict the future because we don't know what will happen in the future.

If you flip a coin and get heads ten times in a row, the odds it will be heads on the next flip are still fifty fifty. Yet the chance that you'll flip a coin eleven times in a row is very low.

If you take a poll it's possible someone just had a bad day. It's unlikely that you randomly caught all people who had a bad day.

A poll can't predict who will be president. There are examples of the polling being wrong. Yet, there are things the poll can tell you that enable you to know the likelihood of someone being president.

The earlier it is, the less you can tell about the outcome of the election for the simple reason that way too many things will change public opinion before the election, the most important of which is that the "swing voters" or as I like to call the, the morons, don't even make up their minds until October. Yet you can tell certain things, like, "No one with these numbers has ever done such and such."

Like in the NFL. Every year you'll get something like, no team who has gone 0 and X has ever made the postseason since, 19XX. Yet someday a team will do that very thing.

The results of one poll amount to a single datum point. The only thing you can discern from that information is exactly and only what it is: the opinions of the people on that date.

Multiple polls provide multiple data points. With that information you can identify trends, correlations, etc.

From the previous polls and the manipulated information gleaned from them, you can apply the conclusions to the single poll and make hypotheses about what one can expect to follow logically from that particular set of results.

It's not unlike the identification of diet components with an individual's health. John's 288 cholesterol doesn't tell you anything more than the fact that John has 288 cholesterol. But the cumulative collection of individual datum points like John's have identified a trend that says he is likely to have heart issues because of the cholesterol. It doesn't say he does have heart issues. It doesn't say he will have heart issues. It simply says he's likely to have heart issues. But his specific results by themselves only tell us he has 288 cholesterol. Everything else comes from the inclusion of additional information independent of John's cholesterol score.

One poll is John's cholesterol score. Several polls create the trend that tells us he's likely to have heart issues because of it.

I never said you couldn't use the results of multiple polls and trending to make logical deductions about the results of a single poll. I said only that the poll itself doesn't actually tell you that. Someone challenged me on this point. And hence the discussion began. However, if you or anyone else can show me how John's cholesterol number by itself tells you anything other than John's cholesterol number, I'm all ears.

 

[melensdad]

Sorry but you are wrong.

The results of one poll that does not use a statistically significant data sample or uses biased questions, or is otherwise faulty yield a flawed result.

The results of one poll that is properly done, with correct sample sizes, etc will yield a result that is greater than the opinion of simply the polled individuals on the polling date.

 

[Sarge48]

Can we just go on and take a poll of who believes in polls and who thinks polls are a bunch of crap?

(Or did I word that wrong? Is it biased towards crap? Must we assume...strike that...presume the folks in this poll () are educated on the subject of polls enough to qualify them to take said poll? )

 

[Sarge48]

But to go on.....

If the poll is done by the media, run, screaming for the hills.
If conducted by a more reliable pollster, then decide for yourself.

 

[dross]

Sorry but you are wrong.

The results of one poll that does not use a statistically significant data sample or uses biased questions, or is otherwise faulty yield a flawed result.

The results of one poll that is properly done, with correct sample sizes, etc will yield a result that is greater than the opinion of simply the polled individuals on the polling date.

I don't think she was saying that you couldn't apply the results to the population the sample represents. I think she's making a distinction about what the poll itself actually shows, and if I understand her correctly, she's right, though I'm not sure how that applies to the larger point.

But if I understand her correctly, I single poll conducted on a single date, conducted correctly shows only what that particular polls shows for that date. In order to determine anything else from the poll you need other information from outside that poll, like other polls taken at other points in time that show a trend.

Which is technically correct, but it's a pretty fine distinction and I'm not sure how it fits into the larger discussion.

 

[dross]

The point I'd like to emphasize on this thread is that dismissing polls out of hand is as silly as automatically believing them.

 

[melensdad]

. . . but it's a pretty fine distinction and I'm not sure how it fits into the larger discussion.

Agreed. But I think she was using her very fine distinction to make a lame attempt at attacking all polling data.

The point I'd like to emphasize on this thread is that dismissing polls out of hand is as silly as automatically believing them.

EXACTLY. As I said before, research the polls, be skeptical but be willing to believe.