tag:blogger.com,1999:blog-19346366.post7318443143198569727..comments2024-03-27T23:28:20.274-04:00Comments on Face to Face: Availability bias in judging young people based on YouTube, MySpace, etc.agnostichttp://www.blogger.com/profile/12967177967469961883noreply@blogger.comBlogger14125tag:blogger.com,1999:blog-19346366.post-41917214678308811372008-11-12T00:06:00.000-05:002008-11-12T00:06:00.000-05:00Intelligence seems to be very relevant to this deb...Intelligence seems to be very relevant to this debate. Trzesniewski used it, in conjunction with self-reported intelligence to measure narcissism. Twenge discusses his methods in this area and criticizes them:<BR/><BR/><I><BR/>Trzesniewski et al. (this issue) examined self-enhancement in the Monitoring the Future data using the residual of the correlation between self-rated intelligence and self-reported high school grades. Even when an objective measure of performance taken at a different time is used, this type of self-enhancement correlates only modestly with narcissism (r=.22; Paulhus & Williams, 2002). It is completely unknown how narcissism correlates with Trzesniewski et al.'s measure, which uses self-reported grades measured at the same time as self-reported intelligence. Even if most students accurately report their grades, those who inflate their reports may be the same ones who inflate their intelligence (Farwell & Wohlwend-Lloyd, 1998). In addition, the MTF survey does not ask for GPA but instead a self-report of "your average grade so far in high school" on a 9-point scale, an even more subjective measure. Thus Trzesniewski et al.'s calculation relies on two subjective measures, both of which are correlated with narcissism (e.g., Farwell & Wohlwend-Lloyd, 1998; in contrast, objective measures of performance are not correlated with narcissism). In consequence, the residual scores are virtually meaningless.<BR/><BR/>Second, self-reported intelligence and grades are four items apart in the same assessment, which encourages consistency between the two responses. Third, the intelligence question asks respondents to compare themselves to others their age, which reduces the opportunity for self-enhancement. Fourth, this is essentially a difference score. As such, it has the usual problems of difference scores: It is difficult to tell if beliefs about one's own intelligence have increased and beliefs about peer intelligence have increased, or both have decreased, or both have stayed constant. Last, self-reported grades have increased substantially over time in the MTF data. Only 18% reported earning an A or A- average in 1976 (M=5.78), compared to 33% in 2006 (M=6.34; d across 30 years=0.29). The number who consider themselves "A students" has thus increased by over 80%. Thus the primary story in the MTF data is one of significant grade inflation (or, at least, self-reported grade inflation)—a clear indicator of a culture of narcissism.</I><BR/><BR/>I don't think *either of them* are taking into account the fact that college IQ's have dropped over time.. which they clearly should be....Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-19346366.post-54861877922921035882008-11-11T17:52:00.000-05:002008-11-11T17:52:00.000-05:00Without being versed in this literature, I can't s...Without being versed in this literature, I can't say a whole lot. <BR/><BR/>But I think it's important to note that we <I>do</I> know that the demographics of college students have changed dramatically over time. The IQ level of the average college student <A HREF="http://inductivist.blogspot.com/2008/06/mean-iq-of-college-grads-dropped-9.html" REL="nofollow">has dropped</A> 2/3s of a standard deviation since the 60s. It is plausible that the people who go to college are different in a lot of ways.<BR/><BR/>In fact the lower IQ might be related to the increase in narcissism. What kind of person with a C average and an IQ of 100, thinks he has what it takes to go to college? This is the equivalent of the skinny nerd trying out for the football team, without any sort of irony.<BR/><BR/>Another thing to check for is measurement invariance. Changes across time could be invalid, which is what we are finding with the Flynn Effect.Jason Malloyhttps://www.blogger.com/profile/04855482153162314172noreply@blogger.comtag:blogger.com,1999:blog-19346366.post-81049397670748365422008-11-11T16:30:00.000-05:002008-11-11T16:30:00.000-05:00About our samples: They were college students from...<I><BR/>About our samples: They were college students from intro psych subject pools – exactly the same source as the data in the Trzesniewski et al. paper. And if anything, ours were more representative of the average college student: we had 85 samples from 31 campuses across the country and they had 8 samples from 3 campuses in one region that also happened to be half Asian-American (most college samples are only 10% Asian-American).<BR/></I><BR/><BR/>There has been an interesting trend in terms of college attendance in the last 20 or 30 years.<BR/><BR/>I wonder at the male/female distribution among intro psych and the IQ distribution.<BR/><BR/>Perhaps all they are measuring is the greater ability of neurotic people to get into college as a result of the gradual dumbing down of college courses.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-19346366.post-71514299738942832452008-11-10T19:29:00.000-05:002008-11-10T19:29:00.000-05:00Except that those studies were not designed to des...Except that those studies were not designed to <B>describe the distribution of a trait in a population</B>, as are polls of, say, political opinions or the prevalence of some disease (where narcissism is the opinion or disease).<BR/><BR/>Therefore, we conclude nothing about any purported change over time in the distribution of the trait, and stick with the null of no change.agnostichttps://www.blogger.com/profile/12967177967469961883noreply@blogger.comtag:blogger.com,1999:blog-19346366.post-38309366695154633532008-11-10T16:41:00.000-05:002008-11-10T16:41:00.000-05:00ben g let me know about this discussion, so I thou...ben g let me know about this discussion, so I thought I’d jump in to clarify a few things.<BR/><BR/>The issue of changing populations over time (e.g., more Asian-Americans in a sample) is completely different from an interaction. Interactions address whether the trend is different for one group or another. If all ethnic groups increase in narcissism, there would be no interaction. But if there are more people of a low-scoring group (say, Asian-Americans) in the sample over time, there would be the false appearance of no change, because every group’s scores were rising but the later samples had more low-scoring people. And there would still be no interaction, just changing sample composition over time.<BR/><BR/>But don’t take my word for it. Look at Trzesniewski et al.’s own data, which we did in this upcoming publication in the Journal of Research in Personality. <BR/><BR/>http://www.sciencedirect.com/science/journal/00926566<BR/><BR/>Go to articles in press (in red on the left) and look at the one that’s currently 29. (Mapping the scale of the narcissism epidemic) <BR/><BR/>In short, their data shows an increase in NPI scores too -- at twice the rate that we found, and all ethnic groups show increases over time.<BR/><BR/>About our samples: They were college students from intro psych subject pools – exactly the same source as the data in the Trzesniewski et al. paper. And if anything, ours were more representative of the average college student: we had 85 samples from 31 campuses across the country and they had 8 samples from 3 campuses in one region that also happened to be half Asian-American (most college samples are only 10% Asian-American). Two of their samples were the same as ours, as we gathered all of the samples we could find who gave the NPI to undergraduates, not just a few as they did. There’s your availability bias.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-19346366.post-62559655187825693022008-11-10T13:22:00.000-05:002008-11-10T13:22:00.000-05:00"We're only seeing these girls in the two-week pha..."We're only seeing these girls in the two-week phase of their menstrual cycle when their testosterone levels skyrocket -- during ovulation, more or less. In this phase, females are more likely to want to go out, to be flirtatious, to dress more provocatively, and so forth. Outside of these two weeks per month, they're more reserved and leave-me-alone (to put it nicely). Dr. Jekyll and Mr. Hyde should have been written about a teenage girl."<BR/><BR/>If there is a God, his design of females to be governed by their hormones at various weeks of the month to alter their fundamental personalities must have been some kind of huge, CRUEL joke on mankind. No wonder the ancients just arranged marriages and didn't take women's thoughts of feelings into account, because those can change from one week to the next and back again all within one month depending on whats gushing on the inside of them. They seem to get over it as they age, but oh those young ones..............they can be like children, going through so many emotions over a span of a day that they have to recover from them with "hard" sleep every night. It must be exhausting not to have an even keel like us men do---the stronger sex.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-19346366.post-91411662148766503252008-11-09T23:55:00.000-05:002008-11-09T23:55:00.000-05:00I dunno. The article says that gender and ethnicit...I dunno. The article says that gender and ethnicity don't moderate the time trend -- neither interaction term accounts for more than 0.1% of the variance in NPI scores.agnostichttps://www.blogger.com/profile/12967177967469961883noreply@blogger.comtag:blogger.com,1999:blog-19346366.post-63080016544042195242008-11-09T22:14:00.000-05:002008-11-09T22:14:00.000-05:00The second article in the footnote came out in Aug...<I>The second article in the footnote came out in August, while that Twenge article came out in May, so that's presumably the rebuttal.</I><BR/><BR/>I think they were published in the same journal in May. So this debate seems not to have settled but perhaps if you wrote a post on gnxp criticizing Twenge's rebuttal, she would come and defend herself. For example, I wonder what her response would be to:<BR/><BR/><I>But the article shows that there was no interaction effect between race and year in predicting narcissism</I>.<BR/> <BR/>She seems to be eager to rebut.. my guess is that she'd have some kind of response to this.. and I think it would be cool to see what she has to say.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-19346366.post-50699970826812520172008-11-09T20:16:00.000-05:002008-11-09T20:16:00.000-05:00In plain English, the problem is that the NPI stud...In plain English, the problem is that the NPI studies were not designed to describe a population -- whether young people in general or just college students.<BR/><BR/>If a study isn't carefully designed to describe a population, the sample doesn't tell you much.<BR/><BR/>In fairness, a lot of other people, probably including me, could have made the same mistake since most psychologists and related people aren't interested in describing populations. That's what pollsters and health monitoring agencies do. Usually we just want to know if variable X is associated with variable Y.agnostichttps://www.blogger.com/profile/12967177967469961883noreply@blogger.comtag:blogger.com,1999:blog-19346366.post-75782715488379367892008-11-09T20:07:00.000-05:002008-11-09T20:07:00.000-05:00From the second cited article:The samples used in ...From the second cited article:<BR/><BR/>The samples used in the Twenge et al. (this issue) meta-analysis<BR/>consisted of college students from conventional 4-year institutions, a<BR/>segment of the population that represents approximately 20% of<BR/>American youth aged 18 to 24 (U.S. Census Bureau, 2005a, Table 9;<BR/>U.S. Census Bureau, 2005b, Table 2). Convenience sampling makes<BR/>it impossible to estimate sampling errors accurately (Pedhazur &<BR/>Schmelkin, 1991, p. 321), so there is no way to evaluate the validity<BR/>of population inferences. As noted by the eminent epidemiologist<BR/>Jane Costello (personal communication, 13 June 2007), ‘‘I entirely<BR/>agree that no population conclusions can be drawn from convenience samples—Epidemiology 101.’’ This same constraint on population<BR/>inferences from convenience samples applies to the meta-analytic<BR/>estimates derived from aggregates of convenience samples. Indeed,<BR/>history has shown that bigger samples are not necessarily better<BR/>samples for making predictions about the behavior of populations of<BR/>interest. Recall that George Gallup used a relatively small but representative<BR/>sample to predict correctly that Roosevelt would win the<BR/>1936 election, whereas the Literary Digest used a much larger, but<BR/>nonrepresentative, sample and incorrectly predicted Landon would<BR/>win that election. In that case, the benefits of a larger sample size<BR/>were not offset by the limitations of a nonrepresentative sampling<BR/>strategy. Thus, any inference about changes in the psychological attributes<BR/>of American youth or even American college students is on<BR/>uncertain grounds when the raw materials are convenience samples<BR/>of college students. It is possible, for example, that the secular trends<BR/>observed in Twenge et al.’s meta-analysis reflect changes in the kinds<BR/>of people who attended college or participated in psychology research<BR/>in the late 1970s and 1980s compared to the 1990s and 2000s.<BR/>To our minds, this sampling issue is the single biggest limitation of<BR/>the Twenge et al. meta-analysis for understanding cohort changes in<BR/>narcissism.<BR/><BR/>[...]<BR/><BR/>In collecting studies to include in their meta-analysis, Twenge et al. (this issue) excluded studies that did not report a mean<BR/>for the NPI in the published article. This strategy of limiting published<BR/>studies to those that reported NPI means might be a reasonable<BR/>approach for creating a manageable sample of studies; however,<BR/>it reduces the available pool of studies and creates potential biases.<BR/>For instance, we published an article (Donnellan, Trzesniewski,<BR/>Robins, Moffitt, & Caspi, 2005) that included one large study with<BR/>NPI data (N53,143), but we omitted the means due to the strict<BR/>space constraints of Psychological Science and their lack of relevance<BR/>to our focal research question. Adding the 3,143 participants from<BR/>our article (M514.87; date coded as 2003) to Twenge et al.’s (this<BR/>issue) analysis (i.e., the means reported in their Table 1) reduces the<BR/>secular increase in NPI scores by almost half, from .53 to .28. Thus,<BR/>the effect reported by Twenge et al. is somewhat volatile if the inclusion<BR/>of one, albeit large, sample can cut the effect size so<BR/>substantially. To be sure, there is no way to determine how the<BR/>Twenge et al. results would ultimately be affected if they had contacted<BR/>all authors of published papers to acquire unreported means;<BR/>however, we believe that this step would have resulted in a data set<BR/>that was larger and more representative of all studies using the NPI<BR/>with college students.agnostichttps://www.blogger.com/profile/12967177967469961883noreply@blogger.comtag:blogger.com,1999:blog-19346366.post-92146460759946652952008-11-09T18:59:00.000-05:002008-11-09T18:59:00.000-05:00The second article in the footnote came out in Aug...The second article in the footnote came out in August, while that Twenge article came out in May, so that's presumably the rebuttal.<BR/><BR/>I don't think most psychologists know anything at all about statistics (even less than I do), and these NPI studies weren't designed to accurately measure current levels of narcissism. They just study whoever shows up for intro psych most of the time.<BR/><BR/>That's why the CDC wants to measure the prevalence of some disease, they go through lots of trouble to ensure their design gets them a representative sample. It doesn't come for free.<BR/><BR/>Twenge says that the large Asian pop at UC Davis explains some of the lack of change in narcissism -- more Asians over time would obscure increases in White narcissism. But the article shows that there was no interaction effect between race and year in predicting narcissism.agnostichttps://www.blogger.com/profile/12967177967469961883noreply@blogger.comtag:blogger.com,1999:blog-19346366.post-20245656949223097682008-11-09T17:24:00.000-05:002008-11-09T17:24:00.000-05:00Hmm, looks like Twenge isn't going down without a ...Hmm, looks like Twenge isn't going down without a fight:<BR/><BR/>http://www3.interscience.wiley.com/journal/120120843/abstract<BR/><BR/>Any thoughts on their rebuttals?Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-19346366.post-28315890975698335922008-11-09T15:58:00.000-05:002008-11-09T15:58:00.000-05:00The data Twenge looked at weren't representative s...The data Twenge looked at weren't representative samples, while the ones showing no change were. To show a change in the prevalence of X in group Y over time, you have to make sure you're sampling the same group Y.<BR/><BR/>They also didn't send out a request for unpublished data from authors who work in the area, which you usually do when you do a meta-analysis. Some data from Trzesniewski weren't published because of space constraints, and that single study reduced Twenge's effect size by a good amount.<BR/><BR/>I don't know that personality isn't changing at all -- but young people are not more narcissistic than in the 1970s. Before that, there is no data (that anyone knows of). Young people sure seemed pretty self-absorbed during the Roaring Twenties, Gay Nineties, and the Protestant Reformation.<BR/><BR/>Over the time-span that most people are talking about -- two or three generations -- there would have to be huge changes to result in genetic changes.<BR/><BR/>As for facultative responses to changing environments, same thing -- society hasn't changed so radically in the past one to two generations. It only looks that way if you restrict the view to the past one to two generations -- rather than the past 500, 3000, 10000, or 1 million years.<BR/><BR/>Also, we're just talking about basic personality traits, not observed behavior. Maybe the level of narcissism is the same among young people today, but technology allows narcissists to more easily broadcast their personality to the world than before.<BR/><BR/>Or levels of criminal personality traits could stay the same, while criminal behavior went through cycles, depending on how the populations of criminals and restrainers interact.agnostichttps://www.blogger.com/profile/12967177967469961883noreply@blogger.comtag:blogger.com,1999:blog-19346366.post-51461406956052779922008-11-09T12:00:00.000-05:002008-11-09T12:00:00.000-05:00Interesting. A couple questions:So is Twenge's th...Interesting. A couple questions:<BR/><BR/>So is Twenge's thesis basically garbage? (Rogler 2002), too?<BR/><BR/>And if that's the case, was McCrae right that personality hasn't changed over the last couple of decades in modern societies?<BR/><BR/>And if we aren't changing at all, personality-wise, why not? Haven't studies provided evidence that there's an environmentaly driven difference between personalities in more primitive societies as compared to our own? Since modern society is becoming increasingly 'modern', how come this difference isn't increasing as time goes on?Anonymousnoreply@blogger.com