meazza Posted November 10, 2006 Share Posted November 10, 2006 look bub, I do stats every friggin' day...taken numerous undergrad and grad level stats courses, as well as experimental design...I can safely say you don't know what your talking about, period, end of story. 831841[/snapback] you clearly have no self-esteem. Link to comment Share on other sites More sharing options...
Wraith Posted November 10, 2006 Share Posted November 10, 2006 BJ, Ramius and myself...there were others...like good old T-Bone (who BTW is confounded by his support of BJ's position in this debate). 831956[/snapback] I've already seen BJ argue, incorrectly, that the behavior HA describes does not exist. It does. I have then seen that same person, BJ, as well as yourself and few others, argue that the phenonemon is caused by the normal distribution of the population and the error. I have also seen that same group argue that HA is arguing that measurement error is causing the behavior to happen. That group has also argued that HA is arguing that measurement error is causing a regression towards the mean. I have seen HA say that measurement error is necessary for this event to happen. I have not seen HA say that measurement error is causing a regression towards the mean. Please show me where HA supposedly made this claim that measurement error causes regression towards the mean. Also, keep in mind that at least one of the published scientists/statisticians has already made an incorrect claim in this thread. So if HA did make the claim, admit that it happens and move on as it is clearly not pertinent to HA's stated hypothesis or the current argument. Most of you are arguing over semantics. Link to comment Share on other sites More sharing options...
Nervous Guy Posted November 10, 2006 Share Posted November 10, 2006 Please show me where HA supposedly made this claim that measurement error causes regression towards the mean. 831979[/snapback] ..."Now suppose that you introduce measurement error into this test. Someone who measured out at 6'2" may actually be a 6'0" person who got lucky on the first test. Because there is now measurement error in the system, those who obtained exceptionally high measurements the first time are likely to regress toward the mean upon being remeasured. " he said it. Link to comment Share on other sites More sharing options...
Nervous Guy Posted November 10, 2006 Share Posted November 10, 2006 you clearly have no self-esteem. 831971[/snapback] none...zero...nadda...zilch. Link to comment Share on other sites More sharing options...
Wraith Posted November 10, 2006 Share Posted November 10, 2006 ..."Now suppose that you introduce measurement error into this test. Someone who measured out at 6'2" may actually be a 6'0" person who got lucky on the first test. Because there is now measurement error in the system, those who obtained exceptionally high measurements the first time are likely to regress toward the mean upon being remeasured. " he said it. 831987[/snapback] Because I have displaced this rubberband, it will now snap back. Absolutely true. Of course elasticity is the cause. In that quote at least, he is not saying the CAUSE is measurement error. He is correct in implying that without measurement error, the regression towards the mean would not happen (for the obvious reason that without measurement error there would be no deviation from the mean in the first place). Semantics again. Link to comment Share on other sites More sharing options...
meazza Posted November 10, 2006 Share Posted November 10, 2006 none...zero...nadda...zilch. 831990[/snapback] There there. Link to comment Share on other sites More sharing options...
GG Posted November 10, 2006 Share Posted November 10, 2006 I've already seen BJ argue, incorrectly, that the behavior HA describes does not exist. It does. I have then seen that same person, BJ, as well as yourself and few others, argue that the phenonemon is caused by the normal distribution of the population and the error. I have also seen that same group argue that HA is arguing that measurement error is causing the behavior to happen. That group has also argued that HA is arguing that measurement error is causing a regression towards the mean. I have seen HA say that measurement error is necessary for this event to happen. I have not seen HA say that measurement error is causing a regression towards the mean. Please show me where HA supposedly made this claim that measurement error causes regression towards the mean. Also, keep in mind that at least one of the published scientists/statisticians has already made an incorrect claim in this thread. So if HA did make the claim, admit that it happens and move on as it is clearly not pertinent to HA's stated hypothesis or the current argument. Most of you are arguing over semantics. 831979[/snapback] Then who said this? You are making your arguments overly complicated. I agree completely that the methodology is shaky. However, you do not even need to look that closely to understand his conclusion is wrong: He is trying to say that by retaking the test, he is showing that error is causing a regression towards the mean. However, by retaking the test, you are mitigating the effects of measurement error. If his methodology truely showed a regression towards the mean, he would be helping to prove the opposite hypothesis of his own. It's simple. Link to comment Share on other sites More sharing options...
Orton's Arm Posted November 10, 2006 Share Posted November 10, 2006 Then who said this? 832016[/snapback] It's possible I should have gone into more detail when explaining how I created my simulation. What I did was as follows: 1. created a population of 1000 members, and assigned each a true I.Q. The true I.Q.s were based on a normal distribution. 2. Gave each population member an I.Q. test based on their true I.Q. plus a random error term. 3. Assigned those members with the highest test scores to the Threshold group. 4. Gave Threshold members a second I.Q. test based on their true I.Q.s plus a random error term. On the second I.Q. test, Threshold members obtained slightly lower scores than on the first I.Q. test. This is because Threshold members were selected in part based on their luck on the first I.Q. test. The second time Threshold members were given the test, good luck presumably balanced bad; resulting in slightly lower scores. Link to comment Share on other sites More sharing options...
Wraith Posted November 10, 2006 Share Posted November 10, 2006 Then who said this? 832016[/snapback] Of course I did. I then asked Holcomb's Arm to clarify his position. He responded to me quickly and publicly: "Perhaps I need to be more specific about what my hypothesis actually is. (Although it's not really "my" hypothesis since I read about it elsewhere.) Someone who gets an extremely high score on an I.Q. test is likely to get a somewhat lower score if that person is retested. Someone who gets a very low score on an I.Q. test is likely to get a slightly higher score if retested. This phenomenon would vanish if there was no measurement error on either of the tests." - HA. Link to comment Share on other sites More sharing options...
syhuang Posted November 10, 2006 Share Posted November 10, 2006 I may have missed it when he explicitely stated that measurement error is causing the regression towards the mean, but that doesn't seem to be what the argument is about here at all anymore. 831941[/snapback] How about this I was measuring whether error in measurement can cause the appearance of regression toward the mean! the presence of an error term causes the appearance of regression toward the mean. It causes the regression toward the mean. or this My model is only designed to say one thing, but to say it very well: error in measurement can cause the appearance of regression toward the mean. or this? Or is it the fact that neither of you supposed statistics experts can understand a very simple concept such as how measurement error can cause the appearance of regression toward the mean? The quotes with 'appearance' may be able to be rationalized by your explanation. But definitely not this one "It causes the regression toward the mean". Link to comment Share on other sites More sharing options...
Orton's Arm Posted November 10, 2006 Share Posted November 10, 2006 How about If you look at those quotes in the context of the examples I was providing, you'll get a better idea as to what I was getting at. Link to comment Share on other sites More sharing options...
GG Posted November 10, 2006 Share Posted November 10, 2006 Of course I did. I then asked Holcomb's Arm to clarify his position. He responded to me quickly and publicly: "Perhaps I need to be more specific about what my hypothesis actually is. (Although it's not really "my" hypothesis since I read about it elsewhere.) Someone who gets an extremely high score on an I.Q. test is likely to get a somewhat lower score if that person is retested. Someone who gets a very low score on an I.Q. test is likely to get a slightly higher score if retested. This phenomenon would vanish if there was no measurement error on either of the tests." - HA. 832035[/snapback] Which to me, would indicate the probability of regressing to the mean in a normal distribution (regardless of measurement error) and HA's penchant for changing the direction of the conversation when caught with error pants down. Perhaps he needs to be more specific the first 10 times he tries to explain his position, then this cluster of a thread wouldn't rival NJ Sue. Link to comment Share on other sites More sharing options...
Wraith Posted November 10, 2006 Share Posted November 10, 2006 The quotes with 'appearance' may be able to be rationalized by your explanation. But definitely not this one "It causes the regression toward the mean". 832045[/snapback] Yep, that one was indeed a poor choice of words. Link to comment Share on other sites More sharing options...
Wraith Posted November 10, 2006 Share Posted November 10, 2006 Which to me, would indicate the probability of regressing to the mean in a normal distribution (regardless of measurement error) and HA's penchant for changing the direction of the conversation when caught with error pants down. Perhaps he needs to be more specific the first 10 times he tries to explain his position, then this cluster of a thread wouldn't rival NJ Sue. 832051[/snapback] Couple of points: - "Regardless of measurement error?" This particular behavior only happens when there is measurement error. - He definitely took a long time to explain his hypothesis in detail but this thread would not have degenerated as it did without the presence of certain other members of this board, who have also let their arguements evolve over time. Link to comment Share on other sites More sharing options...
Orton's Arm Posted November 10, 2006 Share Posted November 10, 2006 Perhaps he needs to be more specific the first 10 times he tries to explain his position, then this cluster of a thread wouldn't rival NJ Sue. Back on page 12, I wrote Consider a world in which the following was true: you had a perfect I.Q. test, which gave you the same score every time. Also in this example, the heritability term in the I.Q. equation is now 1. So far, so good--there is no regression toward the mean. But now, keep everything the same, except that you allow your I.Q. test to be flawed. The underlying truth is that if two parents have an average I.Q. of 140, their children will be expected to have I.Q.s that are the same. But as I illustrated in my earlier post, measurement error creates a distorted view of things. So the people who got 140s on the I.Q. test really have an average I.Q. that's somewhat lower. This is because there are more people with I.Q.s of 130 or 135 (available for getting lucky) than there are people with I.Q.s of 150 or 145 available for getting unlucky. So if you have a group of people who all got a 140 on an I.Q. test, the true average I.Q. of that group will be less than 140. If you give that group a second I.Q. test, its true average will make itself clear. But in this case, the second I.Q. test could take the form of those people having kids. The average child from that group will reflect that group's true (130s-style) average, and not its inflated average of 140. "Look!" people might say, "The kids are regressing toward the mean. Why is this?" The reality is that the kids' I.Q.s are, on average, the same as their parents'. It's just that you overestimated how smart those parents were. Bungee Jumper responded with this: What you have going on here is what was best expressed by Wolfgang Pauli: "That's not right, that's not even wrong!" What he meant was, when it comes to science, it's possible to be right, it's possible to be wrong, or it's possible to be such an ineffable idiot that you haven't even achieved anything resembling science. An excellent example of this is pretending the number "130" is an acceptable substitute for a gaussian distribution. You're not right, you're not even wrong. Wraith is right to say that my choice of words hasn't always been ideal. But it's been very frustrating for me to continuously try to explain a statistical phenomenon, and have people just not get it. And not only did they not get it, they repeatedly ridiculed the explanations I provided. You'll notice, by the way, that my explanation on page 12 says the same thing as my explanations on page 23. I've been saying the exact same thing for the last twelve pages! I'm glad that finally someone's joined this thread who actually gets it. Link to comment Share on other sites More sharing options...
Bungee Jumper Posted November 10, 2006 Share Posted November 10, 2006 BJ, Ramius and myself...there were others...like good old T-Bone (who BTW is confounded by his support of BJ's position in this debate). 831956[/snapback] And myself. Physics Review E for one...that would be the volume concerned with statistical physics... Link to comment Share on other sites More sharing options...
Bungee Jumper Posted November 10, 2006 Share Posted November 10, 2006 Because I have displaced this rubberband, it will now snap back. Absolutely true. Of course elasticity is the cause. In that quote at least, he is not saying the CAUSE is measurement error. He is correct in implying that without measurement error, the regression towards the mean would not happen (for the obvious reason that without measurement error there would be no deviation from the mean in the first place). Semantics again. 831999[/snapback] There is another thread he started about statistics (it may be on page 2 of the board already). Maybe it's there. But he did say that according to a researcher named Weiss, the cause of regression toward the mean was error... Link to comment Share on other sites More sharing options...
Nervous Guy Posted November 10, 2006 Share Posted November 10, 2006 And myself. Physics Review E for one...that would be the volume concerned with statistical physics... 832078[/snapback] uh...you're BJ (AKA CTM, DCT, etc...) Link to comment Share on other sites More sharing options...
Bungee Jumper Posted November 10, 2006 Share Posted November 10, 2006 Of course I did. I then asked Holcomb's Arm to clarify his position. He responded to me quickly and publicly: "Perhaps I need to be more specific about what my hypothesis actually is. (Although it's not really "my" hypothesis since I read about it elsewhere.) Someone who gets an extremely high score on an I.Q. test is likely to get a somewhat lower score if that person is retested. Someone who gets a very low score on an I.Q. test is likely to get a slightly higher score if retested. This phenomenon would vanish if there was no measurement error on either of the tests." - HA. 832035[/snapback] And that's a perfect example of regression toward the mean...of the friggin' error. You take a test, you score X. The test has no error. You take it again, you score X. If the test has error, normally distributed with a standard deviation of s, you have roughly a 68% chance of scoring within s of X, a 95% chance of scoring within 2s of X... Now postulate two different but related tests measuring the same thing, one exact and one with inherent normally distributed error. On the exact test you score X. You take the "errored" one. You score X + 2s. You take it again. There is a 95% chance that you will score within 2s of X...which means there's a 98% chance that you will score less than X + 2s. This is because the error is normally distributed with an associated probability of occurrence. It is not because X is normally distributed. It's not. It's fixed by the first test. The regression toward the mean is the regression toward the mean of the normally distributed paramater - the error. It's as I've been saying; he doesn't know what he's actually measuring. He set up this wacky simulation, measured the regression of a variable, then proceeded assign said regression to the wrong !@#$ing measurable. THAT'S why regression disappears if he eliminates error...because when he eliminates error, he eliminates the regression of such. It's because he can't even begin to understand that he's confusing two different variables that he thinks error is "causing" regression. Link to comment Share on other sites More sharing options...
syhuang Posted November 10, 2006 Share Posted November 10, 2006 There is another thread he started about statistics (it may be on page 2 of the board already). Maybe it's there. But he did say that according to a researcher named Weiss, the cause of regression toward the mean was error... 832080[/snapback] here it is: link 3. The article fits in nicely with what I've said--that intelligence is largely genetic, that parental intelligence drives the intelligence of children, etc. The formula which I cited is basically a mathematical definition of regression toward the mean. I've addressed the issue elsewhere. And at least for Weiss, regression toward the mean is simply an artifact of measurement error, rather than a genuinue phenomenon. Certainly it's not something that should be allowed to stand in the way of an otherwise welcome eugenics program. and link What did I mean when I wrote that regression toward the mean could be caused by error in measurement? Suppose that the expected value of a child's intelligence is the average of the two parents' I.Q.s. Now suppose that the people in the I.Q. 140 room had been kept there so long they started having kids with each other. Based on the I.Q. tests those people took, you'd expect their kids to have I.Q.s of 140. That was the measured I.Q. But the true average I.Q. in that room is lower--perhaps 135. Once those people's children start taking I.Q. tests, those children will get an average score of 135. This is what Weiss meant when he pointed out that measurement error could explain regression toward the mean. Poor choices of words, again? Link to comment Share on other sites More sharing options...
Recommended Posts