Glen, Why would that data concern you when I have already said that it should not work with CR? You are claiming to be a scientist but you present fraudulent tests. Way back in 1986 I explained that the causes should be uniform in their generation of the effect. Normally sampled causes are truncated. They are as crazy as normally distributed cell sizes in ANOVA. But you create a bogus test that is certain to fail and to NOT test CR according to its assumptions. Perfect correlation is a confound. Did you inspect the tables in my 1991 article that showed that after the correlation passes about .95, CR can not work. A few days ago I called you on your fraudlent test and you said I was lying about your motives. What are you motives? What explains you trying to interpret a test that is illegitimate from the start. This is not science. Its is propaganda. If you want to test CR, then do it right. First test the model I use to define the method. See if I am cheating? See if it is as obvious as Gottfried says it is? Then comment on what is implied by the logic and the simulations. Then using reason instead of propaganda and insinuation, explain why you think CR does not work. Then do the simulations, Then try out survey data and tea leaves. The surey data is the problem. We do not know what to make of it. It should not be used as the criterion. Is milae confounded with gasoline conservation policies? Would the conservation policies cause the manufactures to use different sorts of engines, bodies etc? Its tea leaves when the measures are not done well, so that confounds are removed and convenience samples are not allowed to dictate government policies, economics, and what ever you do for a living.
NEXT... when you have bothered to see the obvious, test the boundaries. Cchallenge the logic. Admit the limitations of your tests and explain the reason for contriving them in the manner that you do. Were you even trainned as a scientist? Consider what you have done. You created a test that CR is not SUPPOSED to pass. Its like asking a political opponent if he has ever had passionate thoughts about a woman other than his wife. What ever he says, you can win votes. When CR fails your test, you conclude something along the lines of what you have been saying thus far. Do you know the meaning of the word Sophistry? Let me ask you some questions. Try to answer them. If you can not then explain why. If you can but do not wish to explain, please stay out of this conversation. If you can not be a scientist, then go somewhere else to smirk and show your ignorance. 1. Let x1 and x2 be uniformly sampled random variables. Does it make sense why I say "sampled" instead of distributed? Distributions can be a function of many different things, many of which are circumstantial. When an experimenter collects equal numbers of observations per combination of a factorial anova, she samples uniformly. She does not simply let the randomness of nature decide the distribution. Do you have problem with this idea of scientific sampling for causal inference? Do you acknowledge that many disciplines have progressed by using such sampling? 2. Does it surprise you that larger values of y tend to occur with random pairings of large x1 and large x2 values? Let there be a range for x1 and x2 from 1 to 10. When we pair x values of 5 and 5 we get a y value of 10 (mid range). When we pair x values of 10 and 10 we get a y value of 20 (pretty darn big). When we pair x values of 0 and 0 we get a y value of 0 (tiny). Thus the very large and very small values of y tend to correspond to those random or crossed occassions when both very large and very small values of x1 and x2 are paired with similarly large and small values. This means that in the extremes of y, x1 and x2 are correlated. Big x1 goes with big x2. Little x1 goes with little x2. Thus, ergo, therefore, x1 and x2 are correlated in the extremes of y BECAUSE y is determined by the values of x1 and x2. Gottfried (but no one on semnet at first) thought this was obvious. Do you see it. 3. Do you see that if we combine very different values of x1 and x2, randomly, then their sum will be in the middle of the y range. They will cancel one another out to the midrange of y. Do you see this? Do you see that with a group of x1s and x2s from the middle of y, the x1s and x2s will tend to be different and thus negatively correlated? Do you see this? Do you see there is a logical relationship between the x and y variables. The relationship is asymmetrical. X variables determing y variables. Y variables do not determine X variables but they reveal the x variables that are by definition latent in their y structure. Scholars are both intelligent and educated. There are uneducated but intelligent people. There are also educated and unintelligent people. Scholarship implies both intelligence and education. Neither intelligence or education imply academic scholarship. Intelligence and education are not enough to be a scholar but they are essential. Along with motivation, opportunity, a love for truth, intelligence, education.... scholarship gains its definition... its instantiation.. its causation from the combination of x1 and x2. Do you understand this? Do you understand how unethical it is for the journal of Structural Equation Modeling to accept a paper, to forbid my sharing it for three years, and then to drop it for political reasons is? You feel any duty to complain about this? Do you read books published by Lawrence Erlbaum? Why do you trust them? Bill "Glen Barnett" <[EMAIL PROTECTED]> wrote in message news:almlf7$g34$[EMAIL PROTECTED]... > > Simon, Steve, PhD <[EMAIL PROTECTED]> wrote in message > news:E7AC96207335D411B1E7009027FC284902A9B24F@EXCHANGE2... > > I was out of the office for a week, so I missed out on most of this > > discussion. Can I inquire what happened when Corresponding Regressions (CR) > > was applied to the test data sets that Glen Barnett supplied? > > I will post answers soon, but I want to give William a chance to look first. > > Glen > . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
