I know I am a little late reading emails and tips has moved on to other topics. I think there can be a lively debate about rotated versus non-rotated, correlated vs. non-correlated. I like to think g as a construct (as well as intervening variable) will persist to the extent it guides theory and research.
The reasons I cling to g in my vocabulary: 1. The predictive validity of g loaded tests increases with the complexity of the task, with 77% of rocket scientist (nuclear weapons specialists) success accounted for by g, versus 20 to 30% of lower level military task success predicted (note this is % variance, the r is the square root), Also note that g predicts a surprising chick of variance for even the lower level tasks. 2. _Achievement_ tests are predicted better by g than by school grades 3. Specific aptitude tests add virtually no predictive value for job success above and beyond g 4. Items with high g loadings _look_ like they are measuring a general ability to manipulate symbol e.g. these items have big g loadings: a. Progressive Matrices b. analogies c. series completion d. reasoning: e.g. "Bob is twice as old as his sister, who is now 7. How old will Bob be when his sister is 40?" Answer: 47 (n.b. the math is trivial. But the reasoning is not) >From a validity construct standpoint, g looks like a duck, it quacks like a >duck, it waddles like a duck etc. So that is why I still believe in my heart >etc etc etc. BUT we probably all agree that neurological findings will >ultimately uncover causal mechanisms. Thanks Mike for a thoughtful reply ... ========================== John W. Kulig, Ph.D. Professor of Psychology Coordinator, Psychology Honors Plymouth State University Plymouth NH 03264 ========================== ----- Original Message ----- From: "Mike Palij" <[email protected]> To: "Teaching in the Psychological Sciences (TIPS)" <[email protected]> Cc: "Michael Palij" <[email protected]> Sent: Wednesday, April 9, 2014 8:36:09 PM Subject: Re: [tips] How Intelligent is IQ John, A few points: (1) One way to think about factor analysis is that it is a technique to explain the mathematical basis for correlations, specifically, being able to reduce a correlation matrix to a smaller matrix where the rows and columns are independent (though one may be satisfied with a smaller factor matrix with correlated factors). With the development of confirmatory factor analysis, we can test whether a single factor or multifactor model best accounts for a correlation matrix. Structural equation modeling (SEM) allows for much more complex factor models. That being said, there are a number of problems in getting adequate fit of models to data as well as problems of interpretation. The key point is that this is all in terms of mathematics and does not necessarily have any psychological relevance. (2) The original conception of "g" depended upon getting a large first eigenvalue or factor from an unrotated factor matrix. The model of the data can be said to be X = common variance + error variance where common variance is "g" when talking about measues of mental abilities. However, an unrotated factor matrix can be difficult to interpret because, well, it may not represent the structure in the data. That is, instead of a single factor, several factors may be required. Rotation of the factors while maintaining orthogonal relations among them (i.e., uncorrelated factors) often reveals several factors. Varimax rotation attempts to maximize the loading of a variable on one factor while minimizing the loadings on other factors. This, of course, undermines the whole "general intelligence" position. This was promoted by Thurstone with his primary mental abilities approach and others (e.g., Guilford; Gardner and his "multiple intelligences" approach is a johnny-come-lately) and this would researchers to focus on tests based on several specific factors (e.g., verbal ability, math ability, spatial, etc.) instead of a single factor. (3) If one allows the factors to be correlated after they are extracted, one can come up with a correlation matrix for the factors which can be further factor analyzed. Proponents of "g" would argue that "general intelligence" was a "higher order" factor in order to save their theory -- but now "g" is even further away from the actual data it is supposed to explain. "g" is now a "second order" factor but some situations might have it as a "third order" factor (a large number of variables and subjects are required to get these higher order factors). "g" has strayed far from its original meaning and it is no longer clear what it means, especially when is refering to second-order or higher unitary factors. In this sense, "g" is an artfact because it comes about the continual factor analyses of correlation matrices -- from the original correlation matrix of empirical variables to the derived correlation matrices for factors. If one believes in his/her heart that "g" exists as a meaningful entity, I'm sure that this seems like a reasonable thing to do. If one does not believe in "g", this seems like grasping at straws. (4) This is a special time of year because the different strains of Christianity celebrate Easter at the same time; see: http://www.almanac.com/content/when-easter The main distinction is whether one follows the "modern" Gregorian calendar or the "old" Julian calendar (they are based on lunar cycles and not all have transitioned from the old calendar to the new old; however, Russia was late in making the transition which explains why the October revolution is celebrated in November). Roman Catholics follow the Gregorian calendar while the Orthodox Christians tend to follow the Julian calendar which means that the holidays are usually out of sync. However, Byzantine rite Catholics historically have followed the Julian calendar though in recent years some have changed to observing the holiday on the Gregorian calendar. For some background on this see: http://www.crisismagazine.com/2011/we-are-non-roman-catholics Bringing this back to the original point of this thread, one could use the metaphor that "g" is like the Catholic Church, that is, an overarching conception that governs all sub-units even those that don't recognize its authority because it can be argued that they are all derived from Roman Catholicism. The different varieties of Catholics and Christians are like specific abilities, representing different components that operate in a common system. So, from this perspective, a "g" enthusiast would focus on the importance of Roman Catholicism as driving all forms of Christianity while people who don't care for "g", well, not so much. ;-) Me, I'm a primary abilities kind of guy. ;-) -Mike Palij New York University [email protected] ----- Original Message ----- On Wednesday, April 09, 2014 12:37 PM, John Kulig wrote: Mike I am not sure I get the point about g being an artifact of factor analysis. I realize we can name factors anything we wish. The loadings correlate the sub-tests with the hypothetical/latent variable that we call factor I, II etc .... I also know that there are different methods of factor analysis, and we can get different results, but if guided by theory/common sense and the result is a construct that succinctly summarizes a broad array of empirical findings, then I do not see the artifact. I do know that a factor will emerge when it predicts differences. So (loosely stealing an example from Cronbach/ the pencil is my example) ... a sub-test of vocabulary and a sub-test of pencil sharpening ability will not see a common factor emerge with homogeneous Ss, even though there is a skill common to both - willingness to sit and follow directions. But if we had a more heterogeneous sample of people from very different cultures, a common factor of "willingness" would emerge to predict differences. In the later example, the "willingness" would be a useful construct, label it what you will. As I think about the neurological underpinnings (jumping from one issue to another) it may be the case that there are numerous brain functions common to all tasks, or maybe only some tasks. Like factor analysis, do they predict differences in the population we get our samples from? And - jumping again - I suspect Mike and I are in a small group who celebrate Greek Passover/Easter given his expertise in that area (add our list to the cross-cultural dudes on tips. UNLESS he is simply an expert in very diverse fields - OMG! is that g????). Whatever the case, have a fruitful equinox holiday season! ----- Original Message ----- On Wednesday, April 9, 2014 12:16:17 PM, Mike Palij wrote: On Wed, 09 Apr 2014 07:50:49 -0700, Jim Clark wrote: >Hi > >I'm surprised to see the IQ bashing based on a perhaps simplistic >interpretation of some brain research showing that two different >areas of the brain light up in 16 subjects performing various cognitive >tasks. [snip] I think you miss the point: it is the use of IQ/intelligence/"g" as theoretical concepts for cognitive or brain processing that is being contested. There are alternative theoretical frameworks that can be used but some people feel compelled to use IQ/intelligence/"g". One might prefer a theory that claims that the Flying Spaghetti Monster fills a person's heads with blue fairies that when active give off energy that is detected by neuroimaging techniques (but I'll leave the debunking of neuroscience results to Tips resident neuroscience debunker Scott Lilienfeld ;-). Hence, every thought you have is the result of a busy blue fairy. Now try to falsify that claim. But do so after you show the evidence for virtual particles. ;-) See the following article in Scientific American but also read the comments: http://www.scientificamerican.com/article/are-virtual-particles-rea/ Then take a look at the Physics FAQ on virtual particles: http://www.mat.univie.ac.at/~neum/physfaq/topics/virtual Oh, and I'm glad that no one has shown that the claim that "g" is an artifact of factor analysis is false. ;-) >Somewhat related, there is an interesting interview with Flynn in >the latest Skeptic magazine. Interesting interview but it leaves one wondering why anybody let the Irish immigrate to their country. ;-) -Mike Palij New York University [email protected] --- You are currently subscribed to tips as: [email protected]. To unsubscribe click here: http://fsulist.frostburg.edu/u?id=13338.f659d005276678c0696b7f6beda66454&n=T&l=tips&o=36005 or send a blank email to leave-36005-13338.f659d005276678c0696b7f6beda66...@fsulist.frostburg.edu --- You are currently subscribed to tips as: [email protected]. To unsubscribe click here: http://fsulist.frostburg.edu/u?id=13338.f659d005276678c0696b7f6beda66454&n=T&l=tips&o=36014 or send a blank email to leave-36014-13338.f659d005276678c0696b7f6beda66...@fsulist.frostburg.edu --- You are currently subscribed to tips as: [email protected]. To unsubscribe click here: http://fsulist.frostburg.edu/u?id=13090.68da6e6e5325aa33287ff385b70df5d5&n=T&l=tips&o=36079 or send a blank email to leave-36079-13090.68da6e6e5325aa33287ff385b70df...@fsulist.frostburg.edu
