Monday Highlights

Good morning.

  1. Keep burning-at-the-stake those who confess and whistleblow … and the outcome will be?
  2. My collegiate experience would vindicate that, one caveat, you have to work hard enough at up to going to sleep so that your sleeping self will work on it.
  3. I was going to say, alcohol was involved, but the first comments indicates that was not true? However, impairment remains on the table, I should think.
  4. Liberals and their hard to shake postulates. (note: this isn’t to say that conservatives, or anyone else, doesn’t have hard-to-shake postulates that are wrong, so don’t go whistling up that tree)
  5. A good final send off question. Why?
  6. Word search.
  7. And a book noted.
  8. Make that books noted, here’s another.
  9. So, was the health tax on tanning beds a sly stab at the gym membership -> conservative conversion effect?
  10. Thumbs up on this end.
  11. Today will tell? Uhm, no .. the day that will tell is a Tuesday in early November.
  12. Myth and consequence.
  13. For the cricket race followers.

27 responses to “Monday Highlights

  1. 9.So, was the health tax on tanning beds a sly stab at the gym membership -> conservative conversion effect?

    Don’t know about your parts but I’ve never seen a gym with tanning beds in my area. Tanning shops are always distinct from gyms.

  2. For the cricket race followers.

    LOL, more like for the right-wing bubble followers.

    “The race this year will hinge on turnout, but I’m thinking that my prediction of Romney at 370 or so is understating the outcome. ”

    ROFL. Nate Silver has it at 237 Obama – 206 Romney so far.

    I just hope on election night you look back on this and can make the analogy to all the other ridiculous things you believe that are at odds with all the experts in that field.

  3. (Of course I know better. Nothing will change and you will make the exact same mistake next time around.)

  4. JA,
    So, Mr Silver’s prediction depends on a D+10, D+12, or D+20? What do you think it will be D+x where x=?

    You’re feeling a lot of Democrat enthusiasm? On what do you base your supposition? Polls that assume the same? Or on what?

    What do you suggest I believe in regards to the outcome? Apparently you think that I am “believing” that the outcome will be 370 or greater? Is that it? One what do you base that supposition? I’m curious.

    Let’s see, I looked at the last election and state by state there were variances of between 4-20% in different polls outcome predictions? This is, according to you, a settled science and not a cricket race.

  5. JA,
    You do realize that if this was not a cricket race all the polls would agree within sampling error. They don’t. They don’t agree with each other or the election results.

    What mistake will I make? Gallup has Romney +7, others -2. The sampling error on that is probably under a percent. How’s that work anyhow? You took a “class” in statistics.

    Oh, wait, you believe in the validity of the results of the scientific field of climatology, which apparently has never made a prediction which you can name or point to … just one of the ridiculous things you believe.

  6. Boonton,
    They used to be more common. The little gym I got to even used to have them. The bigger health clubs all used to have them too. 5-10 years ago.

  7. When I was in NC I noticed a huge number of closed tanning salons….I would guess maybe ten years ago there was some type of boomlet in indep. tanning salons.

    In terms of polls, I check the trading markets and most are keeping at 60%+ in favor of Obama. Suppose they could be wrong but we’re talking multiple markets that aren’t as shallow as they used to be.

  8. This is, according to you, a settled science and not a cricket race.

    False dichotomy much?

  9. Isn’t it a weird coincidence how your worldview depends on you disagreeing with experts in every field? And isn’t it lucky that you are so smart that despite having no relevant expertise you are cleverly able to see the flaws that all the professional pollsters and the people who pay them tons of money are completely oblivious to? And in climate science too — how amazing that you are smart enough to see the flaws that are so glaringly obvious that the best scientists in the field are… completely oblivious to?

    If the flaws you think you see are obvious, don’t you think the experts have considered them already? WTF. The hubris is absurd with you.

  10. 4.Liberals and their hard to shake postulates. (note: this isn’t to say that conservatives, or anyone else, doesn’t have hard-to-shake postulates that are wrong, so don’t go whistling up that tree)

    What postulate is really being attacked here? Does anyone really think foreign aid is some type of pet liberal project to promote economic growth? Very little foreign aid is given with the direct goal of promoting economic growth in mind. Consider that the bulk of our aid is in the form of grants to other countries to buy weapons from us. We’ve also given aid when one country has been struck by some striking diaster. Beyond that foreign aid has never been done on any scale relevant enough to an entire economy with the goal of economic growth.

    Would foreign aid cause economic growth? Of course it would. Are you seriously going to claim that the US economy wouldn’t grow by some amount if, say, China and Japan decided to just drop twenty billion dollars a month at random locations over the US? A more realistic example would be NATO and the Marshal Plan which at the end of the day represented hundreds of billions of dollars transferred from the US to Western Europe. Granted the goal was only economic growth but it was pretty aligned with that and Western Europe has indeed done pretty well on the world stage over the last half century or so.

  11. JA,

    Isn’t it a weird coincidence how your worldview depends on you disagreeing with experts in every field?

    Not every field. I disagree with astrologers and their ilk. This is why the “what results has this field produced” is relevant. And, non-linear dynamicists will tell you they have a new field … and that lots of work needs to be done before they have an established science. Too bad politics prevents honesty like that in climate science.

    And isn’t it lucky that you are so smart that despite having no relevant expertise you are cleverly able to see the flaws that all the professional pollsters and the people who pay them tons of money are completely oblivious to?

    Odd that you can’t see those flaws or admit to them. ’cause I’m pretty sure the actual experts will (unlike true believers like yourself).

  12. I don’t believe Climate is ‘non-linear’…..interesting book I read recently, Turing’s Cathedral, which seems to be more about Neuman than Turing. Anyway one of the early applications of the first computer built at Princeton was climate/weather modeling demonstrated the break where weather ends and climate begins. Weather is modelled best by conditions nearby the source both in terms of geography and time and is highly chaotic. In other words, if you want to know if it will rain tomorrow in NYC, it is more useful to know if it’s raining today in Philly or Chicago. Climate is more easily modelled based on energy flows, heat going in versus heat coming out. Hence, you will know winters in the north are colder than summers not by mapping trillions of temperature points around the globe but by looking at the energy hitting the surface in the north versus the energy radiating out (plus oceanic flows and such). Long story short, it makes a lot of sense that the average global temp may be modelled correctly as rising but the implications of that on local areas will be modelled with a lot of error.

    Odd that you can’t see those flaws or admit to them

    I think the problem is not so much your inability to recognize flaws but your habit of making too much hey about them. You are able to find a flaw and harp on it but I don’t think you’re able to intelligently put the flaw in its proper context. Is the flaw important or trivial? If it’s important how important is it? What other data would tell us it’s so important that the entire model should be trashed? Instead you seem to gravitate towards trying to argue by a set of ancedotes that don’t really lead anywhere useful.

  13. You are able to find a flaw and harp on it but I don’t think you’re able to intelligently put the flaw in its proper context.

    This. For example, the concept that people might lie to pollsters. You bring this up as if it completely invalidates the very notion of polling, as if pollsters never considered the idea — when in reality, it’s an issue that is well-known and well-studied by experts.

    In climate, you ask the most obvious questions like, well what if warming isn’t so bad? As if it never occurred to the scientists to consider whether the effects were negative or positive.

    If you were an honest investigator, you would take these questions/flaws that occur to you and investigate what the experts say about them. Instead, you simply throw them out there as if they invalidate the entire field but all the experts were just too biased or too stupid to think of them themselves.

  14. I would refine that criticism to say you don’t seem committed to the flaws…in other words, suppose you’re making the argument warming won’t be so bad. OK but that means you’re not making an argument against CO2 causing warming, you’re just saying the costs will be on balance very mild. But then you’ll jump up and down citing some ancedote of cold weather somewhere or solar warming or something that has nothing to do with a counter-narrative that the entire warming story works up to the cost estimates…(which would be a fair criticism, although one I’d counter with why risk an uncertain cost when dealing with such a massive system?).

    Or people lieing to pollsters. That kind of begs the question why do they lie? Why are the lie’s asymetrical? Why would, say, Romney supporters lie and say they are Obama voters but not the reverse? If the lies are symetrical then wouldn’t the poll still be valid as errors would tend to cancel out?

    Instead I get the sense that the idea is being clung too not for the sake of understanding but only for the sake of holding your ground in the argument. In other words the argument, fact, or meme isn’t being valued in itself for its truthvalue but rather as a means to an end which damages your credibility. It kind is like what a pornographer is to beautiful women rather than the artist. Sure the former will stumble upon them, but something is missing even when they do.

  15. JA,
    That’s because you’re confused (or I’m confusing). “People lie to pollsters” isn’t the flaw. The flaw is that, for example, right now the major polls have a 12% swing in their “predictions” for the upcoming election … with a less than 1% statistical sampling error. This is the flaw. The “lie to pollsters” is a hypothesis for why the polls are crap. The results of the theory/method are crap on a stick. That’s the problem. Why? Well, lots of reasons likely, lying is just part of it. I’ve suggested some hypothesis for why these are crap. I’m not pretending that these problems aren’t things anyone has considered.

    In climate, recall from our discussion, it turned out “what if warming isn’t bad” wasn’t a criticism (like the above, lying to pollsters wasn’t the flaw) … it was a query. The problem in climate is the missing past legacy of good accurate predictions.

    Instead, you simply throw them out there as if they invalidate the entire field but all the experts were just too biased or too stupid to think of them themselves.

    I see, your abililty to misinterpret my questions is me being biased.

    Boonton,
    Climate being non-linear in part is suggested by the graph on which we were discussing. The fluctuations/noise seem to be fractal, i.e., self similar on time scales that is a marker of non-linearity and chaotic regimes. Seasons are not climate, btw. You can map seasons weather changes by energy flows, global mean temperature and climactic fluctuations … you’ll have to demonstrate that you think power/energy flows dominate the noise.

    But then you’ll jump up and down citing some ancedote of cold weather somewhere or solar warming or something that has nothing to do with a counter-narrative that the entire warming story works up to the cost estimates

    I normally quote “cold weather” as proof of something in the context of some yahoo somewhere citing “warm weather” as proof of something, i.e., if you cite warm as proof of something (and think that is meaningful) while at the same time there are other places at the same time which you ignore which have cold, that is a problem with the source. That is the cold weather meme I push occasionally.

    I think the problem is not so much your inability to recognize flaws but your habit of making too much hey about them. I’m sorry. A 12% fluctuation in polling results isn’t a “minor” flaw. It’s a sign the whole theory/method is crap. If physics had a 10% fluctation in experimental results globally of electron charge or mass … the theory that mass/charge of an electron is something you can measure would be in question. Just like the theory that you can predict/measure prior to the choice/event based on polls is something one should question.

  16. Correct me if I’m wrong but haven’t I seen quite a few posts from you mocking the idea that “climate does not equal weather” whose gist seemed to be lots of ancedotal examples of cold weather somehow refute climate warming?

    I think the problem is not so much your inability to recognize flaws but your habit of making too much hey about them. I’m sorry. A 12% fluctuation in polling results isn’t a “minor” flaw.

    Could you go into detail again where we are getting a 12% flucuation in poll results? Are you comparing polls before an election with election results? If that’s the case are you considering:
    1. The time between the poll and the actual election – there can be shifts in voter opinion between those two points in time.

    2. The question of whether ‘likely voters’ = ‘voters’…which I agree polls should be more explicit. a +- 3% error rate applies to the entire population of people who say they are ‘likely voters’ at a certain point in time, not the actual people who really end up voting.

    3. If polls are so useless, the place to really look would be exit polls. Since these are sampling people who just left the voter booth, we have excellent tools to see if hypothesises like ‘people lie to pollsters’ is true or not and we don’t have to worry about whether we are capturing true voters.

    I don’t recall seeing much here since unless you’re doing analysis, the exit polls are only interesting in the few hours between the polls closing and official tallies being published. I do recall hearing reports that exit polls in Florida showed higher Gore voters than the official tally but I don’t recall actual published results.

  17. Boonton,
    I was going from memory based on the top section of this:

    General Election:

    (IBD/TIPP) Romney 43, Obama 47 Obama +4

    (WashTimes/JZ Analytics) Romney 47, Obama 50 Obama +3

    (CBS News) Romney 46, Obama 48 Obama +2

    (ABC News/Wash Post) Romney 48, Obama 49 Obama +1

    (Monmouth/SurveyUSA/Braun) Romney 48, Obama 45 Romney +3

    (Rasmussen Reports) Romney 50, Obama 46 Romney +4

    (Gallup) Romney 51, Obama 45 Romney +6

    +4 -> -6 is a 11 point swing … right (0 is in there)? I’m comparing apples to apples. General election poll predictions. 10-12 point swing => not science => cricket races.

  18. You’re not describing a 12 point swing. Suppose a poll predicts a 50-50 split with plus/minus 3 points. Missing from that is the confidence interval but let’s assume it’s 95% because that’s pretty standard in stats. All that means if out of 100 times the poll will give you that result, 95 of them will be between 47 and 53 points for any candidate. Nonetheless the average does matter. A candidate getting 49% rather than 50% is more likely than 47%. In other words even within the plus minus 3 points, the likelyhood is that the true number is closer to the sample average than not.

    So say a poll says 48%-52% with +-3 points. You can indeed see the final result being 51%-49% or 44%-55% and that’s within +- 3 points of the poll, even that result should be somewhat surprising in any given election (but not in the lifetime of a pollster who may do hundreds of polls or not in an election season with hundreds of different races).

    To see this more clearly it may be helpful to restate the confidence intervals. If 95% is +-3points 90% may be +- 1.0 points. That would mean in a race where the poll says 49-51, nine times out of ten the second guy will win the race if the poll is accurate.

    Granted this is not as good as Newtonian physics, but it does matter.

    This leads to the second question, given rational actors why pay so much to polling companies if the results are so totally useless? It’s not just the cost of the polling….the Romney campaign has more or less given up on PA, the Obama campaign has likely written off North Carolina. This is all based on their polling data….if the polling is useless they could achieve an easy win by ignoring the polls and putting resources into those ‘non-swing states’ to take them.

  19. Boonton,
    Yes. It is a 12 point swing. And the error bars are likely smaller than 3% (3% is like 6k samples … 2/(sqrt(n)). If you’re taking a 50 state poll, you are likely polling many thousands of people (1k people is 20 per state … which is basically meaningless … and a 6% sampling error).

    The “point swing” is between maxima not the distance between confidence intervals. If the confidence intervals do not overlap (say in particle physics and mass/charge ration estimates) your results are different … and as far as I am aware the difference is expressed as the spread between maxima.

    Instead of expressing the result as 48-52, use a single number (Obama -4 or Obama +4). It makes expressing the result clearer.

    This leads to the second question, given rational actors why pay so much to polling companies if the results are so totally useless?

    Given rational actors (that’s a hoot)! Do you recall the point of the policeman/drunk/light-post joke? That’s the reason.

  20. Instead of expressing the result as 48-52, use a single number (Obama -4 or Obama +4). It makes expressing the result clearer.

    OK express it like that by ignoring the 2nd half of the result. A poll says Obama will get 48% plus or minus 3%. That’s a 6 point wide swing but given that it’s still pretty useful. I think you’re getting tripped up on the ‘magic number’ of 50% which wins the state for Obama if he crosses. But that magic number is arbitrary, in a 3 person race it would be 33%, a ten person race it would be 10%. If you had a 3 person race and Obama was 48% plus or minus 3 points in a state that state would be a clear Obama state and such a poll would be quite useful.

    Given rational actors (that’s a hoot)! Do you recall the point of the policeman/drunk/light-post joke? That’s the reason.

    Politicians are quite rational when it comes to running for office, don’t confuse matters with by the marketing. “Cooco for coco puff’s” isn’t a rational statement but the company that spends millions on that slogan is being quite rational. Are you saying that Romney and Obama are both foolishing spending their limited funds in states like Ohio, Florida, Nevada etc. when they are missing other states that might be legitimately ‘in play’ and provide either one the winning margin with a modest effort by their campaigns?

    The “point swing” is between maxima not the distance between confidence intervals. If the confidence intervals do not overlap (say in particle physics and mass/charge ration estimates) your results are different … and as far as I am aware the difference is expressed as the spread between maxima.

    So let me try to get my mind around this. Say a poll gives 48-52 with an interval of 4 points. Assume that 3rd party votes will be so small that we can assume they will amount to 0%.

    Obama 48% min 44% max 52%

    Romney 52% min 48% max 56%

    This would tell us that in 100 races where the poll gives us this result (think maybe counties), in 95 of them we could expect a list of results that include the two extreme cases of Obama 44, Romney 56 and Obama 52% Romney 48%. But in those 95 cases, the average will still be 48-52 so while 95 is a majority of 100, the end cases would still be a minority inside that majority. We also would know to expect 5 cases where the results will be outside our confidence….say a surprise Obama 55% or a Romney 60%. But the confidence interval is simply something choosen by us. You can make it narrower by allowing more error….so a 90% interval would still capture 90 out of 100 results correctly but vary by less than 4 points, a 99% interval may vary by more than 4 points but will capture more possible results.

  21. Boonton,

    OK express it like that by ignoring the 2nd half of the result.

    And

    Assume that 3rd party votes will be so small that we can assume they will amount to 0%.

    This is why only one number is needed. When you’re told the sum is 100, saying X+4 gives you both haves. Saying x+4 (error 2%) is full information.

    So, say you have a 2% error. One person saying O+4 and another O-7 means there is a 11 point swing and assuming the spread is gaussian (over your samples), these two curves have maximums 11 points apart with errors bars 2% wide. These measurements are statistically worlds apart. If these were mass/charge rations we’d say they were different particles. No overlap. This is the problem. On the one hand, you guys keep claiming this is all good results and not just wild ass guesses. Yet you have well reputed well trained experts giving different results. Your claims that results are methodologically sound are false. This is an “art”, not a science. As an art, you have “good” artists and “bad” ones. This is why people who read the polls will judge different polling agencies as being more or less reliable. Nobody says CERN, Fermilab, and SLC have differing “reliability” ratings regarding their results. One is science. One not, ergo Cricket race is an appropriate categorization.

    People can make a living betting on horses. That doesn’t mean betting on horses is a science or that you should believe any given yahoo who tells you to put your money on the Blue Galoot in the 3rd.

  22. So, say you have a 2% error. One person saying O+4 and another O-7 means there is a 11 point swing and assuming the spread is gaussian (over your samples), these two curves have maximums 11 points apart with errors bars 2% wide.

    Why are you adding together the error bars of two different polls? And no overlap?

    Say one poll says 51% and the other says 52%. If the first is 4 points then it’s range is 47-55. If the other is 7 points then its range is 45-59. To total swing between them is indeed 14 points but that’s not how you would evaluate two polls together. You would combine their samples into a larger sample and calculate a new average and new error range.

  23. Boonton,
    (a) there is no good way, given the information that the polling agencies give of weighting their different contributions.
    (b) Given that many polls do not overlap with error bars … there are clearly methodological differences (because you can’t have “different” particles being measured). Seeing that there are different methodologies … which one is “right”? Why take data from bad methods? How do you factor that in?
    (c) Recall, when I looked at state-by-state on the last election some poll predictions were off by more than 20% from the expected result. Why does this not bother you? Why is this just not like two polls today having spreads greater than their error bar. You do realize this shouldn’t happen at all if the methods were good. If two people measure the same thing in an experiment accurately and they correctly calculate their error bars, a disagreement outside of error is a problem. Yet it is not for polls. Why?

  24. (c) Recall, when I looked at state-by-state on the last election some poll predictions were off by more than 20% from the expected result. ?

    You mean you had a poll in a state that gave Obama 51% when his actual count was 40% or 60%?

  25. Boonton,
    Yes.

  26. (b) Given that many polls do not overlap with error bars … there are clearly methodological differences (because you can’t have “different” particles being measured). Seeing that there are different methodologies … which one is “right”?

    You are drawing cards from a standard deck of 52 and get 4 aces. From that ‘sample’ you conclude the deck has nothing but aces. Someone else draws a sample of 4 and gets number cards, he concludes the deck is mostly or all number cards. Both samples are clearly different but both are correct.

    Given an entire population of, say, 1 million people who split 51%-49% in their votes, it’s perfectly possible to draw a sample of 60 voters one way and 40 voters the other. The error bar is associated with a given level of confidence. When one poll says plus or minus 4 points and another says plus or minus seven the prime driver is either the sample size or the desired confidence interval (90%, 95%, 99% etc.)

  27. Boonton,
    60 voters. 1/sqrt(N) = 1/8th -> 12% error … or +/- 24% at 95% confidence. Your error bars include just about any possible result. Two polls whose error bars do not overlap are inconsistent. Period.

    When one poll says plus or minus 4 points and another says plus or minus seven the prime driver

    Yes and any polls that does not give its error should be ignored. Is that what you’re saying. So … are you suggesting that we are correct to cite any poll that does not include that data as a cricket race (or worse)?

    2 sigma of a Gauss/Normal distribution is assumed as confidence I think unless otherwise specified. Isn’t that right? Sometimes 1 sigma is the normal error. However what it is should be stated.

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>