JFern's "Statistics"

[○∙◄☻¥tπ[╪AV┼cVê└]

Here, since you'll probably reflexly say what I said was wrong since you hate me, you can go here to read about it.

http://www.ruf.rice.edu/~lane/hyperstat/A6642.html

Now let the null hypothesis be that the coin is fair, 50% of the time it gives heads. We collect a random sample of 1000 coin flipping events , where 940 were heads, 60 are tails. We choose our probability of rejecting the null hypothesis to be 0.1%. Do we reject the null hypothesis and conclude that 940 heads and 60 tails is statistically significantly different? Yes or no? You must state either no, and why we can't reject the null hypothesis, or yes, that you finally realized that you're completely wrong.

[J. J.]

I notice you didn't respond to my post on the steps you make to get statistical significance.

The percentage of the time you falsely conclude that there is statisitical significance of something that isn't true is at most p of the time, which is at most 1 in 20 for p=5%. Just because it happens to be the 1 in 20 poll where I falsely conclude statistical significance when it's not true doesn't mean that the poll isn't statisically significant.

About the approximately 1.96 standard deviations for the 95% confidence interval for the normal distribution, if you had  1. looked at the source for that ARG calculator or 2. looked at where I calculated that 2 standard deviations gives 95.44%, using the normal distribution density function of  1/sqrt(2*Pi) * e^(-x^2/2), you'd realize you were wrong. 

I never said that events in July 1933 happened before events in March 1933, you g sh**thead.

Now you still haven't responded to discussion of basic hypothesis testing ffrom freshman statistics. I'd suggest you take that course, since you don't know g sh**t about statistics.

You have set up the wrong hypothesis.  You do not assume that both candidates are at 50%.  The hypothesis is that the poll accurately describes the population; that is where the confidence level comes in.

BTW:  That's why you can get accurate results with 3rd and 4th party candidates that don't get large percentages of the votes.

Boy, you are really bad at this.

[○∙◄☻¥tπ[╪AV┼cVê└]

I notice you didn't respond to my post on the steps you make to get statistical significance.

The percentage of the time you falsely conclude that there is statisitical significance of something that isn't true is at most p of the time, which is at most 1 in 20 for p=5%. Just because it happens to be the 1 in 20 poll where I falsely conclude statistical significance when it's not true doesn't mean that the poll isn't statisically significant.

About the approximately 1.96 standard deviations for the 95% confidence interval for the normal distribution, if you had  1. looked at the source for that ARG calculator or 2. looked at where I calculated that 2 standard deviations gives 95.44%, using the normal distribution density function of  1/sqrt(2*Pi) * e^(-x^2/2), you'd realize you were wrong. 

I never said that events in July 1933 happened before events in March 1933, you g sh**thead.

Now you still haven't responded to discussion of basic hypothesis testing ffrom freshman statistics. I'd suggest you take that course, since you don't know g sh**t about statistics.

You have set up the wrong hypothesis.  You do not assume that both candidates are at 50%.  The hypothesis is that the poll accurately describes the population; that is where the confidence level comes in.

BTW:  That's why you can get accurate results with 3rd and 4th party candidates that don't get large percentages of the votes.

Boy, you are really bad at this.

I was assuming a binary choice. 94% Kerry or heads, 6% Bush or tails, and 0 for any other choice.

Yet again J.J misses the g point because he's a g moron.

Typical, focus on the minuta, where you'er wrong anyways, and ignore the big points I was making, and completely ignore my asking you if you undertand the basic testing of null hypothesis in statistics.

You are a typical Republican, you are a ing moron and a fraud.

[J. J.]

I never said that events in July 1933 happened before events in March 1933, you g sh**thead.

Here is what I said exactly:

  Don't worry, this isn't as bad claiming that something that happened in July of 1933 triggered something in March of 1933.  :-)  That was last week.

You did make that claim, and I get it and post it here.

[○∙◄☻¥tπ[╪AV┼cVê└]

Let's simplify this to questions that require a simple yes or no answer

1. For a sample of 1000, is the 95% confidence MOE closer to 1.96 standard deviations or 2.00 standard deviations?

2. What about for the normal distribution?

3. Suppose our null hypothesis is that a coin is heads 50% of the time. Our false reject p=0.1%. Can we reject the null hypothesis if our sample is 940 heads and 60 tails?

4. Is the probability of getting 940 or more heads, or 940 or tails less than or greater than 0.1%?

The answers to these questions will give great insight to where your lack of understanding is.

[J. J.]

I was assuming a binary choice. 94% Kerry or heads, 6% Bush or tails, and 0 for any other choice.

Yet again J.J misses the g point because he's a g moron.

Typical, focus on the minuta, where you'er wrong anyways, and ignore the big points I was making, and completely ignore my asking you if you undertand the basic testing of null hypothesis in statistics.

You are a typical Republican, you are a g moron and a fraud.

The confidence level does not measure a binary choice; it is the measure ultimately of the accuracy of the sample.  You are really not bright enough to understand that.

Here is the actual commentary; that was Tedrick's point BTW.

You must be logged in to read this quote.

http://www.surveysystem.com/sscalc.htm

You know, like most of your claims, readers can click the link and determine the accuracy.  Ironically, you posted the link in the first place.

[○∙◄☻¥tπ[╪AV┼cVê└]

I was assuming a binary choice. 94% Kerry or heads, 6% Bush or tails, and 0 for any other choice.

Yet again J.J misses the g point because he's a g moron.

Typical, focus on the minuta, where you'er wrong anyways, and ignore the big points I was making, and completely ignore my asking you if you undertand the basic testing of null hypothesis in statistics.

You are a typical Republican, you are a g moron and a fraud.

The confidence level does not measure a binary choice; it is the measure ultimately of the accuracy of the sample.  You are really not bright enough to understand that.

Here is the actual commentary; that was Tedrick's point BTW.

You must be logged in to read this quote.

http://www.surveysystem.com/sscalc.htm

You know, like most of your claims, readers can click the link and determine the accuracy.  Ironically, you posted the link in the first place.

There's only two choices. How ing hard is this for your sorry ass who shouldn't have passed freshman statistics to understand? As for what Tredrick said, if you had any intellect, you'd realize it's not contradictory to what I've said.

[○∙◄☻¥tπ[╪AV┼cVê└]

Again, what would it take for you to ever admit you're wrong?

Can you answer those questions

Choose 95 or 99.9% confidence

1. Is 940 heads, 60 tails statistically significantly different from that of a fair coin?
2. Is a random sample of 1000 with Kerry 94%, Bush 6% statistically significantly different from that of a fair coin
3. Sample of 1000 people, is the 95% MOE closer to 1.96 standard deviations than 2.00?
4. Normal distribution, is the 95% MOE closer to 1.96 standard deviations than 2.00?

If you understand anything about statistics, you should be able to answer yes to all 4 in less than a minute.

[J. J.]

Let's simplify this to questions that require a simple yes or no answer

1. For a sample of 1000, is the 95% confidence MOE closer to 1.96 standard deviations or 2.00 standard deviations?

2. What about for the normal distribution?

3. Suppose our null hypothesis is that a coin is heads 50% of the time. Our false reject p=0.1%. Can we reject the null hypothesis if our sample is 940 heads and 60 tails?

4. Is the probability of getting 940 or more heads, or 940 or tails less than or greater than 0.1%?

The answers to these questions will give great insight to where your lack of understanding is.

The first question is gobblygook.  As has been pointed out, repeatedly, the MOE is not even solely determined SD.  You indicated that here:

2. For large samples, and even larger populations, the normal distribution gives a very good approximation, and for the normal distribution the 95% confidence interval is 1.96 standard deviations in either direction. Of course the MOE varies as the percentage, because if you knew basic statistics, you'd know that the standard deviation is sqrt(p*(1-p)/n), which obviously varies with the percentage.

I'm going to enter your fantasy land until you admit that MOE is not related in that way to SD.  Starting with a false premise will only lead to more false premises.

What you can say in that MOE will increase with a higher confindece level, a smaller sample size, and the closer the score is to the median.

It is interesting how you ask for a yes or no answer and then give an "or" choice.

[J. J.]

Again, what would it take for you to ever admit you're wrong?

Can you answer those questions

Choose 95 or 99.9% confidence


2. Is a random sample of 1000 with Kerry 94%, Bush 6% statistically significantly different from that of a fair coin
3. Sample of 1000 people, is the 95% MOE closer to 1.96 standard deviations than 2.00?
4. Normal distribution, is the 95% MOE closer to 1.96 standard deviations than 2.00?

If you understand anything about statistics, you should be able to answer yes to all 4 in less than a minute.

First, I am only going to answer those related to your initial question.

2.  No, because for one thing, MOE will come into play, because the poll is sample of a larger group.  When you deal with samples, things change.

3-4. are more of your nonsense; you don't get 95% MOE.  You might be referring to confidence levels, but they are not the same thing. 

[ATFFL]

jfern, your knowledge of theoretical statistics cannot be doubted, but your knowledge of applied statistics is lacking in this case.

You are assuming that the polls are measuring random variables.  Who people vote for is not random.  They do not take a perfect coin and flip it to decide who to vote for.  People decide based on a vast number of variables that, even in a dead even 50-50 race, is not something that can be perfectly measured.

If we were arguing about coin flips, you would be spot on perfect correct about nearly everything you have said.  Sadly for you, we are not talking about coin flips.

This is why pollsters need to check validity.  For a coin tossing experiment all we need to do is make sure we a have a perfect coin, one weighted so neither heads nor tails is biased and to make sure we do not bias the toss by not properly spinning the coin when we toss it.

For polls of people we have to run a lot more checks.  We need to make sure our sample is sufficiently random.  Then we need to make sure our random sample is representative of the population as a whole.  Then we need to make sure our questions and readers were unbiased.  Then we need to run our results through a "smell test" to see if they are obvious outliers compared to other results from a similar time frame.  If there are no similar results we have to decide if the numbers are so obviously bad releasing them will hurt our reputation as a pollster. 

Let’s say we get that poll showing Kerry up 96-4.  Is that a statistically significant lead?  Of course it is.  Does that mean that Kerry is leading by that much?  Of course not.  We still need to check if the poll is valid by seeing if we have a random sample, a representative sample and is not an obvious outlier. 

If the electorate is tied 50-50 and we get that 96-4 result we know that our result is garbage and need to toss it.  Even if it is a representative random sample showing a massive, statically significant lead, it is going to fail the last test.  When every pollster shows the race at or near 50-50 but us, odds are we are the ones blowing it. 

Now if we do a poll of African Americans who are registered democrats living in Washington D.C. that 96-4 result begins to look a lot better. 

If our polling was done in a theoretically perfect world where voters decided how to vote based on perfectly random means you would be spot on correct.  We do not live in that world.

[○∙◄☻¥tπ[╪AV┼cVê└]

jfern, your knowledge of theoretical statistics cannot be doubted, but your knowledge of applied statistics is lacking in this case.

You are assuming that the polls are measuring random variables.  Who people vote for is not random.  They do not take a perfect coin and flip it to decide who to vote for.  People decide based on a vast number of variables that, even in a dead even 50-50 race, is not something that can be perfectly measured.

If we were arguing about coin flips, you would be spot on perfect correct about nearly everything you have said.  Sadly for you, we are not talking about coin flips.

This is why pollsters need to check validity.  For a coin tossing experiment all we need to do is make sure we a have a perfect coin, one weighted so neither heads nor tails is biased and to make sure we do not bias the toss by not properly spinning the coin when we toss it.

For polls of people we have to run a lot more checks.  We need to make sure our sample is sufficiently random.  Then we need to make sure our random sample is representative of the population as a whole.  Then we need to make sure our questions and readers were unbiased.  Then we need to run our results through a "smell test" to see if they are obvious outliers compared to other results from a similar time frame.  If there are no similar results we have to decide if the numbers are so obviously bad releasing them will hurt our reputation as a pollster. 

Let’s say we get that poll showing Kerry up 96-4.  Is that a statistically significant lead?  Of course it is.  Does that mean that Kerry is leading by that much?  Of course not.  We still need to check if the poll is valid by seeing if we have a random sample, a representative sample and is not an obvious outlier. 

If the electorate is tied 50-50 and we get that 96-4 result we know that our result is garbage and need to toss it.  Even if it is a representative random sample showing a massive, statically significant lead, it is going to fail the last test.  When every pollster shows the race at or near 50-50 but us, odds are we are the ones blowing it. 

Now if we do a poll of African Americans who are registered democrats living in Washington D.C. that 96-4 result begins to look a lot better. 

If our polling was done in a theoretically perfect world where voters decided how to vote based on perfectly random means you would be spot on correct.  We do not live in that world.

For the coin flipping, it's very obvious what the answer should be. If you have a truly random with replacement sample of people who will only say they support Bush or Kerry, it's equivalent. Obviously the 2nd example isn't realistic, but that's not the point.

This isn't a debate about randomness, so think of the coin tossing. experiment.
 
Yes, in the real world, a 94% Kerry 6% Bush poll is obviously wrong, and one has to worry about all sorts of systematic errors in the polling process.

I was trying to make more mathematical arguments, and not worry about the fact that women with 2 year olds are less likely to answer pollersters during dinner time, or whatever.

[○∙◄☻¥tπ[╪AV┼cVê└]

J.J, I win.

Anyone who understands statistics will agree with the following 2 statements.

1. The radius of the 95% confidence interval (the MOE) for a sample of 1000 or the normal distribution is approximately 1.96 standard deviations

2. Assuming any reasonble probability p for falsely rejecting the null hypothesis, if we get 940 heads and 60 tails in a sample 1000, we must reject the null hypothesis of a fair coin, and conclude we are have statistically significantly difference from a fair coin.

In order to be consistant, you have to disagree with those statements.

[J. J.]

There's only two choices. How g hard is this for your sorry ass who shouldn't have passed freshman statistics to understand? As for what Tredrick said, if you had any intellect, you'd realize it's not contradictory to what I've said.

The confidence level is not the measure of two choices.  It is the liklihood that poll itself represents the population; it is a measure of the accuracy of the sample.  You just don't understand that:

Once again, here is the description;;try to read and comprehend it, especially the bolded parts:

You must be logged in to read this quote.

http://www.surveysystem.com/sscalc.htm

[○∙◄☻¥tπ[╪AV┼cVê└]

Still waiting for you to respond to this, rather then missing the point of what I was saying in some other argument.

I want it to be clear to anyone who undestands statistics, that you're really dense.

J.J, I win.

Anyone who understands statistics will agree with the following 2 statements.

1. The radius of the 95% confidence interval (the MOE) for a sample of 1000 or the normal distribution is approximately 1.96 standard deviations

2. Assuming any reasonble probability p for falsely rejecting the null hypothesis, if we get 940 heads and 60 tails in a sample 1000, we must reject the null hypothesis of a fair coin, and conclude we are have statistically significantly difference from a fair coin.

In order to be consistant, you have to disagree with those statements.

[J. J.]

J.J, I win.

Anyone who understands statistics will agree with the following 2 statements.

1. The radius of the 95% confidence interval (the MOE) for a sample of 1000 or the normal distribution is approximately 1.96 standard deviations

2. Assuming any reasonble probability p for falsely rejecting the null hypothesis, if we get 940 heads and 60 tails in a sample 1000, we must reject the null hypothesis of a fair coin, and conclude we are have statistically significantly difference from a fair coin.

In order to be consistant, you have to disagree with those statements.

No I win, because you've just claimed that MOE is the confidence interval, which it isn't.

Try reading:

You must be logged in to read this quote.

http://www.surveysystem.com/sscalc.htm

Anybody interested can click the link and determine that you are not only wrong, but that you have a reading comprehension problem.

2.  As Tedrick has just pointed out, you cannot use the random acts to simulate this.  You also have failed to realize that a poll is a sample where the coint toss is the whole population.

Now, in you little fantasy world where rule where you're JFRAUD the Dense, you may think you've won.  Unfortunately, that isn't reality.

[○∙◄☻¥tπ[╪AV┼cVê└]

J.J, I win.

Anyone who understands statistics will agree with the following 2 statements.

1. The radius of the 95% confidence interval (the MOE) for a sample of 1000 or the normal distribution is approximately 1.96 standard deviations

2. Assuming any reasonble probability p for falsely rejecting the null hypothesis, if we get 940 heads and 60 tails in a sample 1000, we must reject the null hypothesis of a fair coin, and conclude we are have statistically significantly difference from a fair coin.

In order to be consistant, you have to disagree with those statements.

No I win, because you've just claimed that MOE is the confidence interval, which it isn't.

Try reading:

You must be logged in to read this quote.

http://www.surveysystem.com/sscalc.htm

Anybody interested can click the link and determine that you are not only wrong, but that you have a reading comprehension problem.

2.  As Tedrick has just pointed out, you cannot use the random acts to simulate this.  You also have failed to realize that a poll is a sample where the coint toss is the whole population.

Now, in you little fantasy world where rule where you're JFRAUD the Dense, you may think you've won.  Unfortunately, that isn't reality.

I asked for yes or no answers. 

1. The MOE is the radius of the confidence interval, increase the massive list of J.J making mistake and then saying I made the mistake by 1.

2. I'm talking about a coin toss. It's damn well completely random, you moron, increase the J.J. mistakes list again.

In J.J. reality, the Obama Keyes race was statistically tied, along with Senator Byrd's 2000 race, and other such nail-biters.

I rest my case.

[J. J.]

1. The MOE is the radius of the confidence interval, increase the massive list of J.J making mistake and then saying I made the mistake by 1.

2. I'm talking about a coin toss. It's damn well completely random, you moron, increase the J.J. mistakes list again.

In J.J. reality, the Obama Keyes race was statistically tied, along with Senator Byrd's 2000 race, and other such nail-biters.

You can quit, but on the first one, what relevance does this have to the validity of the poll and specifically to the Confidence Level?

On the second one, tossing 1000 coins is not a random sample of a population.  What makes you think it is J[FRAUD?

[Alcon]

JFern, your earlier comments were way over the line. Please watch it.

[J. J.]

I'd just be happy if he'd answer the question.

[○∙◄☻¥tπ[╪AV┼cVê└]

JFern, your earlier comments were way over the line. Please watch it.

Why? This guy has been calling me JFraud for months, when he's the one wrong. I take offense to the fact that you warned me and not him.

[Citizen James]

Boldly going where angels fear to tread....

Firstly, a coin flip is a poor analogy for political polling.   With a coin flip the odds are known.

The most frequently used analogy is a jar full of colored marbles.  Say some are red and some are blue.   If you pull a number out at random, you can generally get a decent estimate of what percentage total are of each color.

There are of course caviats to that.   Your sample has to be truely random (which is one of the things that makes real life polling so tricky), and it's always possible that you might happen on a rare occurance.  If the marbles are, say, 60% red and 40% blue you can reasonably assume that if you draw out 200 that about 120 will be red and 80 will be blue.  That will not always be exact, more often it will be more like 110-130 red and 70-90 blue.  It is possible, though unlikely, that you might end up pulling out 190 red and 10 blue.  (I beleive this is what JJ is trying to point out).

This is where it gets tricky.  In dealing with the theoretical marbles, we're assuming you know the actual percentages.  With people you don't automatically assume.   If you get results you don't expect you have to weigh at least three options: 1) your polling methods were flawed, 2) your poll was an outlier, or 3) your original premise was incorrect, and the actual percentages were different than what you assumed.

Of course, for the blatently partisan, they will tend to claim that any poll, or batch of polls, that say something they don't want to hear are outliers, no matter how much data piles up.  (the tobacco companies kept up this front for decades)   I suspect this is what Jfern is trying to say.

In any case, Isn't all the name calling back and forth a bit childish?   Calling someone else a fraud just because you don't agree with their position tends to reflect rather poorly on the maturity of the attacker IMO.

[Alcon]

JFern, your earlier comments were way over the line. Please watch it.

Why? This guy has been calling me JFraud for months, when he's the one wrong. I take offense to the fact that you warned me and not him.

Feel free to call him a rather silly nickname constantly, but direct personal attacks are beyond.

[○∙◄☻¥tπ[╪AV┼cVê└]

Boldly going where angels fear to tread....

Firstly, a coin flip is a poor analogy for political polling.   With a coin flip the odds are known.

The most frequently used analogy is a jar full of colored marbles.  Say some are red and some are blue.   If you pull a number out at random, you can generally get a decent estimate of what percentage total are of each color.

There are of course caviats to that.   Your sample has to be truely random (which is one of the things that makes real life polling so tricky), and it's always possible that you might happen on a rare occurance.  If the marbles are, say, 60% red and 40% blue you can reasonably assume that if you draw out 200 that about 120 will be red and 80 will be blue.  That will not always be exact, more often it will be more like 110-130 red and 70-90 blue.  It is possible, though unlikely, that you might end up pulling out 190 red and 10 blue.  (I beleive this is what JJ is trying to point out).

This is where it gets tricky.  In dealing with the theoretical marbles, we're assuming you know the actual percentages.  With people you don't automatically assume.   If you get results you don't expect you have to weigh at least three options: 1) your polling methods were flawed, 2) your poll was an outlier, or 3) your original premise was incorrect, and the actual percentages were different than what you assumed.

Of course, for the blatently partisan, they will tend to claim that any poll, or batch of polls, that say something they don't want to hear are outliers, no matter how much data piles up.  (the tobacco companies kept up this front for decades)   I suspect this is what Jfern is trying to say.

In any case, Isn't all the name calling back and forth a bit childish?   Calling someone else a fraud just because you don't agree with their position tends to reflect rather poorly on the maturity of the attacker IMO.

The argument was not over polling methodology, the argument was over statistics, particularly, setting up a null hypothesis that you have a fair coin (or that Bush and Kerry each have 50%). You then use your poll of coin tosses (or a random sample of Bush and Kerry supporters, with replacement). Either way you have the same essential problem.

You pick a level of significance p, and then take a sample, and determine whether you reject the null hypothesis, and conclude that there is a statistically significant difference.

J.J. doesn't undestand that you can conclude statistical significance from just one poll. He also doesn't understand that for large sample or the normal distribution, the MOE is about 1.96 standard deviations for a 95% confidence interval.

[J. J.]

Boldly going where angels fear to tread....

Firstly, a coin flip is a poor analogy for political polling.   With a coin flip the odds are known.

The most frequently used analogy is a jar full of colored marbles.  Say some are red and some are blue.   If you pull a number out at random, you can generally get a decent estimate of what percentage total are of each color.

There are of course caviats to that.   Your sample has to be truely random (which is one of the things that makes real life polling so tricky), and it's always possible that you might happen on a rare occurance.  If the marbles are, say, 60% red and 40% blue you can reasonably assume that if you draw out 200 that about 120 will be red and 80 will be blue.  That will not always be exact, more often it will be more like 110-130 red and 70-90 blue.  It is possible, though unlikely, that you might end up pulling out 190 red and 10 blue.  (I beleive this is what JJ is trying to point out).

This is where it gets tricky.  In dealing with the theoretical marbles, we're assuming you know the actual percentages.  With people you don't automatically assume.   If you get results you don't expect you have to weigh at least three options: 1) your polling methods were flawed, 2) your poll was an outlier, or 3) your original premise was incorrect, and the actual percentages were different than what you assumed.

Of course, for the blatently partisan, they will tend to claim that any poll, or batch of polls, that say something they don't want to hear are outliers, no matter how much data piles up.  (the tobacco companies kept up this front for decades)   I suspect this is what Jfern is trying to say.

In any case, Isn't all the name calling back and forth a bit childish?   Calling someone else a fraud just because you don't agree with their position tends to reflect rather poorly on the maturity of the attacker IMO.

The argument was not over polling methodology, the argument was over statistics, particularly, setting up a null hypothesis that you have a fair coin (or that Bush and Kerry each have 50%). You then use your poll of coin tosses (or a random sample of Bush and Kerry supporters, with replacement). Either way you have the same essential problem.

You pick a level of significance p, and then take a sample, and determine whether you reject the null hypothesis, and conclude that there is a statistically significant difference.

James42 has it right and JFRAUD has it wrong again.  First, the coin toss is NOT a sample of a large group.  Second, the model here is not that the two candidates (or more) are tied.  It is if the poll itself repesents the greater population accurately (within the MOE).  There is no way, using the internal numbers in the poll, to determine that.  Randomly, the sample can be bad. 

We can say that at at the confidence level assigned there in a certain likelihood that the result is accurate to within a certain range of numbers, the MOE. 

The MOE will shrink with as the sample size increases and when the score moves away from the median; it will grow larger when the confidence level increases.

JFRAUD cannot seem to conprehend that.  He also cannot seem to understand that the MOE has no relationship with if the poll itself is accurate or is an "outlier."

You must be logged in to read this quote.

This of course has no relevence to if the poll is an outrider or not.  It has no effect on the confidence level.  A poll can be properly constructed, and properly conducted, with a score outside of the MOE and it will still NOT be an accurate repesentation of the population.