John McArdle - US grants

The Wechsler Adult Intelligence Scale (WAIS) has been widely used in research as well as in clinical applications. Much of what is believed to be known about adulthood intellectual development is based on research with the WAIS. The conclusions that can be drawn from this WAIS research are often clouded and sometimes contradictory. The results have mixed implications in regard to important questions concerning the factorial structure, developmental sequences and changes, and, in general, the construct validity of the WAIS. During the past two years WAIS data has been collected from surveys of the research literature and from direct contact with researchers and clinics. WAIS summary data, subscale scores, item data and other information on over 40,000 individual subjects has been obtained from over 100 different experimental studies. These data have been coded and stored in a computerized databank and are being used in a large-scale mega-analysis of the WAIS. Initial analyses of these data has focused on the subpopulation of 13,632 healthy adults ranging in age 15 to 95 from 41 different USA experiments. Probabilistic and statistical stratification procedures have been used to select specialized subsamples of interest. Cross validation strategies based on multistage split sampling have been used to examine alternative substantive ideas about the factorial structure of WAIS abilities. A variety of latent-structure path analysis (cf. LISREL) models have been used to study the multivariate dynamics underlying factor growth curves for age, year of testing, and education. Analytic and empirical studies of factorial invariance over age have been initiated and preliminary results recorted. Additional WAIS data (N=7,000) will be collected from known sources and added to the databank. A large scale comparison among the different Wechsler forms (WAIS, WB-I, WB-II, WAIS-R, and WISC-R) will provide an extensive and up-to-date scale equating for a much broader age and cohort range (N=28,000). The available longitudinal data (N=2,000) will be analyzed in terms of cross-lagged factor influences, by comparison to cross-sectional trends, and for subject attrition sampling biases. Latent structure aging trends will also be examined in the context of available data on clinical diagnostic categories (N=12,000), cross-cultural groups (N=5,000), and behavior genetic relative groups (N=3,000).

The Wechsler Adult Intelligence Scale (WAIS) has been widely used in research as well as clinical applications. But the conclusions about adult intelligence that can be drawn from the existing WAIS literature are often clouded and sometimes contradictory. WAIS summary data, subscale scores, item data and other information on over 50,000 individual subjects has been obtained from over 200 different experimental studies. Statistical stratification and crossvalidation procedures have been used to select specialized subsamples for analyses. This research aims is to define and isolate the growth curves of adult intelligence as measured by the WAIS. Large scale Longitudinal data (N=2,500) and family data (N=3,500) have already been collected from many diverse studies, and ranging over the entire adult age span. The larger sample of cross- sectional WAIS data (N=16,000) will be used to match individuals based on some crucial demographic characteristics. Statistical comparisons of the longitudinal, family, and cross-sectional WAIS data will be made to quantify test-retest and within-family selection effects. A variety of latent variable structural equation models (e.g., LISREL, etc.) will be used to examine the basic questions of factorial invariance over age and time. Growth curve models will be developed which will account for components of age, date of testing, education, and gender, as well as subject selectively and biases due to attrition. One aspect of this work will revolve around extensions of the latent growth model (LGM), especially for use with the analysis of incomplete growth curves, and other analyses will use graphical techniques. In a broad sense, we will adapt contemporary methods from human physical growth and epidemiology for use in psychological aging research.

Our previous research on the Wechsler Adult Intelligence Scale (WAIS; NIA grants AG02695, and AG 07407) has produced: (1) a large-scale data base of WAIS information useful for aging research, (2) some novel structural equation models appropriate for aging research, (3) some complex results about the measurement basis of the WAIS, and (4) some complex results about the Growth Curves of Intellectual Ability. One clear result is that the WAIS provides good measures of Crystallized Intelligence, but it does nor provide a good measurement of Fluid Intelligence. Another result is that Growth and Decline processes, are not a simple function of age, but mst be understood in terms of other individual characteristics as well. In this continuation proposal we provide an experimental and efficient plan to unravel some of this complexity in adult abilities research. We propose to test a highly selected set of adults who have already been tested before on a wider battery of ability measures. We will recruit and test N=300 adults from our available files who have (1) been tested before on the WAIS or the Woodcock-Johnson (WJ) tests of intellectual ability (2) are between the ages of 19 and 80 years of age, (3) are spread out in geographical areas of the USA, and (4) who have a specific time lag between their tests (i.e., 2 years to 34 years). We also propose to use a more extensive battery of ability tests, including: (5) the WAIS, (6) the Woodcock-Johnson Revised (WJ-R) test of abilities, (7) Horn's additional markers of the main ability factors, and (8) a demographic questionnaire. This is a continuation of our previous work on several ways. First, we developed the questions from substantive issues about the aging of abilities. Second, we have developed the mathematical and statistical models for growth and change y trying to deal with the massive set of complex WAIS data. Third, we have already gathered longitudinal retest data on the Bradway Longitudinal Data and the National Growth and Change Study (NGCS) samples. In contrast to our prior work, these retested subjects and variables are chosen by a structural experimental design that will allow us to use latent variable structural equation models to unravel (1) the effects of retest from growth, (2) the factors of measurement from the factors of growth, and (3) the direct from the indirect aging effects on these factors.

Our previous research on the Wechsler Adult Intelligence Scale (WAIS; NIA grants AG02695, and AG 07407) has produced: (1) a large-scale data base of WAIS information useful for aging research, (2) some novel structural equation models appropriate for aging research, (3) some complex results about the measurement basis of the WAIS, and (4) some complex results about the Growth Curves of Intellectual Ability. One clear result is that the WAIS provides good measures of Crystallized Intelligence, but it does nor provide a good measurement of Fluid Intelligence. Another result is that Growth and Decline processes, are not a simple function of age, but mst be understood in terms of other individual characteristics as well. In this continuation proposal we provide an experimental and efficient plan to unravel some of this complexity in adult abilities research. We propose to test a highly selected set of adults who have already been tested before on a wider battery of ability measures. We will recruit and test N=300 adults from our available files who have (1) been tested before on the WAIS or the Woodcock-Johnson (WJ) tests of intellectual ability (2) are between the ages of 19 and 80 years of age, (3) are spread out in geographical areas of the USA, and (4) who have a specific time lag between their tests (i.e., 2 years to 34 years). We also propose to use a more extensive battery of ability tests, including: (5) the WAIS, (6) the Woodcock-Johnson Revised (WJ-R) test of abilities, (7) Horn's additional markers of the main ability factors, and (8) a demographic questionnaire. This is a continuation of our previous work on several ways. First, we developed the questions from substantive issues about the aging of abilities. Second, we have developed the mathematical and statistical models for growth and change y trying to deal with the massive set of complex WAIS data. Third, we have already gathered longitudinal retest data on the Bradway Longitudinal Data and the National Growth and Change Study (NGCS) samples. In contrast to our prior work, these retested subjects and variables are chosen by a structural experimental design that will allow us to use latent variable structural equation models to unravel (1) the effects of retest from growth, (2) the factors of measurement from the factors of growth, and (3) the direct from the indirect aging effects on these factors.

Our previous research on the Wechsler Adult Intelligence Scale has produced: (1) a large-scale data base of WAIS information useful for aging research, (2) some novel structural equation models appropriate for aging research, (3) some results about the complex measurement functions of the WAIS, and (4) some results about the complexity of growth curve functions of intellectual ability. During the past four years of this project we have measured a strategically selected set of adults who had been tested before on a wider battery of ability measures, including (5) three-wave longitudinal retest data on the National Growth and Change Study (NGCS) sample, (6) a seventh-wave of longitudinal data on the smaller Bradway sample, and (7) current measurements of the Berkeley Growth Study participants. We have also (8) developed dynamic models for unraveling patterns of leading and lagging indicators of aging processes. In this competing continuation proposal we will continue to use and develop new statistical methods for the synthesis of research on the measurement of dynamic changes in the growth curves of intellectual abilities. We will concentrate on the use of the existing databanks to provide an evaluation of the structural, kinematic, and dynamic hypotheses of the "theory of fluid and crystallized intelligence", and we collect new data from newly completed neurological and experimental studies. The analyses planned include (1) improved ability measurements based on item response theory and improved factorial structure through convergent operations, (2) a formal evaluation of the growth and declines of abilities through latent growth and bear dynamic models, (3) a formal evaluation of the age-based leading and lagging indicators of multivariate developmental sequences through some new multivariate dynamic structural equation models, (4) the creation of a new archive of experimental and neurological measures to evaluate convergent measurement relationships using confirmatory modeling techniques, and (5) provide new methods for further research on aging. These new results will help synthesize all past research on the WAIS, explicitly define the relationships of the WAIS with currently collected measures, and clarify the growth and decline of intellectual abilities across the adult life-span.

The Survey Research Center at the Institute for Social Research at the University of Michigan will continue data collection activities for this longitudinal study and produce datasets in co{lat}oration with the University of Southern California for public release. The purpose of this study is to continue longitudinal data collection activities to assess a variety of cognitive abilities using different modes of administration. In particular, we will conduct follow-up interviews with participants in two samples - Cognition and Aging in the USA (CogUSA) and the National Growth and Change Studies (CogNGCS). For this project, we propose to conduct measurements of the CogUSA participants on up to four separate occasions and CogNGCS participants on two occasions using both telephone and internet modes of administration (i.e., a total of >6,800 tests during this project period). Telephone and internet survey instruments will be developed in coliaboration with John J. McArdle at the University of Southern California. In addition, we will collect saliva samples from these participants for genetic material analysis. Data will be analyzed to examine data quality prior to delivering the data to the University of Southem California research team. These data will be used to create a large, longitudinal public-use dataset that will be analyzed and used to understand age trends in cognitive abilities, the measurement of different cognitive abilities in multiple modes of administration, the relationship of cognitive abilities and trajectories to genetic characteristics, and relations to economic behavior indicators, based on a continued collaboration with the Cognitive Economics (CogEcon) project team directed by Robert J. Willis, In addition, these data may be used to inform HRS Co-Investigators about how to improve upon existing HRS cognitive measures in future waves of the HRS, particularly in relation to the validity of cognitive testing via the telephone and internet.