Full Information Maximum Likelihood Estimation Research Articles

Mid-life occupational grade and quality of life following retirement: a 16-year follow-up of the French GAZEL study

Objectives: This article aims to contribute to the literature on life course influences upon quality of life by examining pathways linking social position in middle age to quality of life following retirement in French men and women.Method: Data are from the GAZEL cohort study of employees at the French national gas and electricity company. A finely grained measure of occupational grade in 1989 was obtained from company records. Annual self-completion questionnaires provided information on quality of life in 2005, measured with the CASP-19 scale, and on participants’ recent circumstances 2002–2005: mental health, physical functioning, wealth, social status, neighbourhood characteristics, social support and social participation. Path analysis using full information maximum likelihood estimation was performed on 11,293 retired participants.Results: Higher occupational grade in 1989 was associated, in a graded relationship, with better quality of life 16 years later. This association was accounted for by individuals’ more recent circumstances, particularly their social status, mental health, physical functioning and wealth.Conclusion: The graded relationship between occupational grade in mid-life and quality of life after labour market exit was largely accounted for by more recent socio-economic circumstances and state of health. The results support a pathway model for the development of social disparities in quality of life, in which earlier social position shapes individual circumstances in later life.

Revisiting Aschauer via ICM estimation

In this article, the effect of public expenditure on consumers’ welfare is considered by estimating the substitutability of public expenditure for private consumption in the context of Aschauer (1985). Aschauer’s full information maximum likelihood estimation requires much more assumptions than the equilibrium condition of his model. Therefore, this study proposes an alternative estimation approach under a weaker condition, which is equivalent to the equilibrium condition of the model. In particular, I propose to use an integrated conditional moments estimation and apply the proposed estimation to US economy. The estimation results show that Aschauer’s model specification is not rejected at conventional significance levels, even though substitutability of public expenditure for private consumption is not significantly different from zero.

An Examination of Teachers’ Perceptions of Principal Support for Change and Teachers’ Collaboration and Communication around Literacy Instruction in Reading First Schools

Background/Context Little research has directly examined whether principal leadership can increase the degree to which teachers work together regularly in focused ways around content. Prior research has shown that reform efforts seeking to alter the process of teaching can be successful if teachers collaborate to build capacity and improve instruction. Furthermore, the research literature has highlighted supportive principal leadership as a key component of teachers’ perception of an effective collaborative change process. Purpose This study examines the empirical link between teachers’ perceptions of principal support for change and teachers’ reports of the degree of collaboration and communication with one another around literacy. Setting The data for this study was collected in all 165 Reading First schools in Michigan in the 2006-2007 school year. Subjects Survey data was collected from 1,738 teachers across all schools. Research Design Data for this study was collected as part of the evaluation of Reading First in Michigan. Literacy teachers in kindergarten through third grade completed surveys at three points throughout the school year (fall, winter, and spring). These surveys contained questions about principal support for change and collaboration and communication around literacy. Data Collection and Analysis The primary analytic method employed in this study was multilevel modeling. Factor analysis and full information maximum likelihood estimation were also used. Findings/Results The measure of principal support for change was a significant positive predictor of teachers’ assessment of the degree of regular collaboration and communication around reading instruction; a one standard deviation increase in teachers’ reports of principal support for change was associated with a 0.202 standard deviation increase in teachers’ assessment of the degree of collaboration and communication around literacy when controlling for the other variables in the model. Conclusions/Recommendations This paper demonstrates that the degree to which teachers report collaboration and communication around literacy in Reading First schools is significantly related to their perceptions of principal support for change. If policymakers expect teachers to collaborate around issues of instruction, then they need to consider the principal's role in supporting change in the school by encouraging teachers to improve their instruction and take the risks associated with innovation. In sum, the results of this study are important for educators, as the role of the principal is potentially critical for positive changes in teacher collaboration and, consequently, student achievement in high-poverty schools.

Five Methods to Score the Teacher Observation of Classroom Adaptation Checklist and to Examine Group Differences

This study compared 5 scoring methods in terms of their statistical assumptions. They were then used to score the Teacher Observation of Classroom Adaptation Checklist, a measure consisting of 3 subscales and 21 Likert-type items. The 5 methods used were (a) sum/average scores of items, (b) latent factor scores with continuous indicators, (c) latent factor scores with ordered categorical indicators using the mean- and variance-adjusted weighted least squares estimation method, (d) latent factor scores with ordered categorical indicators using the full information maximum likelihood estimation method, and (e) multidimensional graded response model using the Bock-Aitkin expectation-maximization estimation procedure. Measurement invariance between gender groups and between free/reduced-price lunch status groups was evaluated with the second, third, fourth, and fifth methods. Group mean differences based on the 5 methods were calculated and compared.

Abstract P429: Parallel Processes of Sedentary Behavior and Cardiometabolic Risk Factors in Obese Adolescents

Introduction: Sedentary time has been negatively associated with cardiometabolic health in adults. These associations are less clear in youth, may differ by sedentary measurement method, and rarely have been investigated longitudinally. Hypothesis: Greater accelerometer measured sedentary time and reported screen time will be associated with poorer status on cardiometabolic risk factors at baseline and longitudinally. Methods: Obese (BMI ≥ 95th percentile for age and gender) 11-13 year olds were enrolled in a primary care weight loss trial. Half were girls and 82% were Hispanic. Of the 106 participants enrolled, 73 completed baseline and 12-month assessments. Assessments occurred in the clinic and included a fasting blood draw. Table 1 lists the risk factor variables. The first dependent variable was reported screen time (television and computer/video games). The second was total sedentary time, calculated from waist-worn Actigraph accelerometers as percent of wear time with counts < 100 per 60-second epoch. Mplus software was used to fit latent growth parallel processes models, with the intercept of the biomarker regressed on the sedentary time intercept (cross-sectional association), and the slope of the biomarker regressed on the sedentary time slope (longitudinal association). Models were adjusted for potential confounders. Full information maximum likelihood estimation was used to account for missing data. Results: Participants reported 3.2 (SD = 2.0) screen time hours. Associations between screen time and risk factors are presented in Table 1. Total sedentary time (M = 61±11% of wear time) was positively associated with BMI (β = .22; p < .05) at baseline but not with other risk factors. Total sedentary time was not associated with risk factors longitudinally. Conclusion: Changes in screen time but not total sedentary time were associated with cardiometabolic risk factors, specifically BMI, blood glucose, LDL, AST, and ALT. Associations were independent of physical activity and psychosocial diet variables.

When outcome heterogeneously matters for selection: a generalized selection correction estimator

The classical Heckman (1976, 1979) selection correction estimator (heckit) is misspecified and inconsistent, if an interaction of the outcome variable with an explanatory variable matters for selection. To address this specification problem, a full information maximum likelihood (FIML) estimator and a simple two-step estimator are developed. Monte Carlo (MC) simulations illustrate that the bias of the ordinary heckit estimator is removed by these generalized estimation procedures. Along with OLS and ordinary heckit, we apply these estimators to data from a randomized trial that evaluates the effectiveness of financial incentives for reducing obesity. Estimation results indicate that the choice of the estimation procedure clearly matters.

Return Volatility and Asymmetric News of Computer Industry stocks in Tehran Stock Exchange (TEX)

According to leverage and volatility feedback effects there are relationships between the return and the risk of stocks in the stock markets. Using daily and weekly data of Computer industry index in Tehran stock Exchange (TEX), this study investigates both leverage and volatility feedback effects applying GARCH family models and Full Information Maximum Likelihood (FIML) estimation method, during 01/2007- 10/2013 period. According to GARCH-M model estimations the first hypothesis of the research (Return volatility of computer industry in TEX affects the return significantly) cannot be rejected for daily data during 02/2010 to 10/2013 (the 2nd period) which both return and return volatility were much more volatile rather than 01/2007 - 02/2010 (the 1st period), but this hypothesis can be rejected for daily data in the 1st period and weekly data in both periods. According to EGARCH and TGARCH estimations the second main hypothesis of the research (a negative return makes return volatility of computer industry in TEX more volatile) cannot be rejected for both daily and weekly data in the 1st period, but can be rejected for both data during the 2nd period.

Coresidence With Husband’s Parents, Labor Supply, and Duration to First Birth

This article investigates the time to first birth, treating coresidence with husband's parents and labor supply as endogenous and using representative data on Taiwanese married women born during 1933-1968. We use a full-information maximum likelihood estimator for a duration model with endogenous binary variables. Results controlling for endogeneity suggest that both coresidence and working are associated with a delay in childbearing, reversing the effect of coresidence on the timing of first birth but not that of working. Women in earlier cohorts tend to choose coresidency and not working, and an increasing number of women from later cohorts choose to do both or to work only.

Planned Missing Data Design: Through Intended Missing Data Make Research More Effective

Missing data is a very common phenomenon in social science research. Among most existing statistical analysis about missing data, Full-information maximum likelihood estimation and multiple imputation are recommended as most common approach at present for handling missing data. Planned missing data designs use special research design which create completely random missing data, and then employ modern missing data estimation techniques to get unbiased parameters and maximize statistical power in the process. Planned missing data designs can be used in cross-section survey to increase questionnaire items or reduce respondent burden. In longitudinal survey to shorten measure occasion, moreover, enrich validities. There are two types of planned missing data designs: 3-form design and two-method measurement design.

A Simultaneous Equation Approach on the Relationship between Implied, Realised and Historical Volatility

This article investigates the relationship between implied, realised and historical volatility by taking into account measurement errors. In this study, call and put implied volatility is calculated from June 2001 to April 2012, using non-overlapping 131 monthly at-the-money samples. The study shows analyses of predictive power of call and put implied volatility separately using simultaneous equation approach. The present work also accounts for the error-in-variable problem which is controlled using instrumental variable technique. Two-stage least squares (2SLS) and three-stage least squares (3SLS) estimations have been performed in order to obtain the consistent estimate of implied volatility. Instrumental variable estimation procedure uncovers that implied volatility has been measured with error. Indeed, the study also accounts for the endogenous nature of the call and put implied volatility dealing with full information maximum likelihood (FIML) estimation of the entire system. Ordinary least squares (OLS) result shows that call and put implied volatility best subsumes the information about the future volatility. Option prices (implied volatility) are the best forecast for the realised return volatility and useful in the pricing of risky financial assets.

Modeling Differential Item Functioning Using a Generalization of the Multiple-Group Bifactor Model

The authors present a generalization of the multiple-group bifactor model that extends the classical bifactor model for categorical outcomes by relaxing the typical assumption of independence of the specific dimensions. In addition to the means and variances of all dimensions, the correlations among the specific dimensions are allowed to differ between groups. By including group-specific difficulty parameters, the model can be used to assess differential item functioning (DIF) for testlet-based tests. The model encompasses various item response models for polytomous data by allowing for different link functions, and it includes testlet and second-order models as special cases. Importantly, by assuming that the testlet dimensions are conditionally independent given the general dimension, the authors show, using a graphical model framework, that the integration over all latent variables can be carried out through a sequence of computations in two-dimensional subspaces, making full-information maximum likelihood estimation feasible for high-dimensional problems and large datasets. The importance of relaxing the orthogonality assumption and allowing for a different covariance structure of the dimensions for each group is demonstrated in the context of the assessment of DIF. Through a simulation study, it is shown that ignoring between-group differences in the structure of the multivariate latent space can result in substantially biased estimates of DIF.

Sexual quality of life in Hodgkin Lymphoma: a longitudinal analysis by the German Hodgkin Study Group

Background:Health-related quality of life (HRQoL) comprises different domains of physical, mental, and social well-being. In this analysis, we focus on sexual quality of life in Hodgkin Lymphoma (HL) patients.Methods:Four-thousand one-hundred and sixty patients enroled in the HD10–HD12 trials underwent HRQoL assessment. Instruments included the Quality of Life Questionnaire for survivors (QLQ-S), combining the European Organisation for Research and Treatment of Cancer QLQ-C30, Multidimensional fatigue (FA) inventory (MFI-20) and an additional sexual functioning (SX) scale. We describe SX up to 27 months after therapy and analyse relationship to stage, age, gender, FA, social functioning, and therapy. Statistical methods range from descriptive statistics to a classification of SX courses, and a longitudinal structural equations model with full information maximum likelihood estimation of missing data. In the analysis, a score below 50 was used to describe severe sexual dysfunction.Results:Three-thousand two-hundred and eight patients provided data on SX. Patients in advanced stages reported lower SX than patients in early stages both, before and after the treatment. During follow-up, an improvement of SX compared with baseline was detected, except for those ⩾50 years. Patients in early stages reached normal SX, whereas advanced-stage patients remained below the reference value for healthy controls. Sexual functioning during follow-up was significantly and strongly related to previous SX, other HRQoL measures, age, and stage, and to lesser degree with gender and chemotherapy.Conclusion:Overall, HL patients have a decreased sexual quality of life at baseline, which improves after therapy and normalises in early-stage patients. Importantly, long-term SX is more closely related to patient characteristics and SX at baseline than to the intensity of treatment.

In Search of Synergy Effects: Mergers and Productivity

We propose an alternative method for investigating whether firms improve performance through mergers after taking into account the selection bias of merging firms. We simultaneously consider the dynamics of firm performance and the merger decision by employing full information maximum likelihood (FIML) estimation. Our study differs from previous studies in that state dependence, unobservable heterogeneity, and selection bias are incorporated simultaneously. Because the effects of mergers may be felt gradually, the dynamic effects of mergers and the factors associated with these dynamics should be taken into account. Our FIML approach complements the strategy used in the extant literature for investigating the effects of mergers on firm performance.

The development of delinquency during adolescence: a comparison of missing data techniques

Conclusions on the development of delinquent behaviour during the life-course can only be made with longitudinal data, which is regularly gained by repeated interviews of the same respondents. Missing data are a problem for the analysis of delinquent behaviour during the life-course shown with data from an adolescents’ four-wave panel. In this article two alternative techniques to cope with missing data are used: full information maximum likelihood estimation and multiple imputation. Both methods allow one to consider all available data (including adolescents with missing information on some variables) in order to estimate the development of delinquency. We demonstrate that self-reported delinquency is systematically underestimated with listwise deletion (LD) of missing data. Further, LD results in false conclusions on gender and school specific differences of the age–crime relationship. In the final discussion some hints are given for further methods to deal with bias in panel data affected by the missing process.

Pairwise likelihood estimation for factor analysis models with ordinal data

A pairwise maximum likelihood (PML) estimation method is developed for factor analysis models with ordinal data and fitted both in an exploratory and confirmatory set-up. The performance of the method is studied via simulations and comparisons with full information maximum likelihood (FIML) and three-stage limited information estimation methods, namely the robust unweighted least squares (3S-RULS) and robust diagonally weighted least squares (3S-RDWLS). The advantage of PML over FIML is mainly computational. Unlike PML estimation, the computational complexity of FIML estimation increases either with the number of factors or with the number of observed variables depending on the model formulation. Contrary to 3S-RULS and 3S-RDWLS estimation, PML estimates of all model parameters are obtained simultaneously and the PML method does not require the estimation of a weight matrix for the computation of correct standard errors. The simulation study on the performance of PML estimates and estimated asymptotic standard errors investigates the effect of different model and sample sizes. The bias and mean squared error of PML estimates and their standard errors are found to be small in all experimental conditions and decreasing with increasing sample size. Moreover, the PML estimates and their standard errors are found to be very close to those of FIML.

Individual influence on model selection.

Researchers in psychology are increasingly using model selection strategies to decide among competing models, rather than evaluating the fit of a given model in isolation. However, such interest in model selection outpaces an awareness that one or a few cases can have disproportionate impact on the model ranking. Though case influence on the fit of a single model in isolation has been often studied, case influence on model selection results is greatly underappreciated in psychology. This article introduces the issue of case influence on model selection and proposes 3 influence diagnostics for commonly used selection indices: the chi-square difference test, Bayesian information criterion, and Akaike's information criterion. These 3 diagnostics can be obtained simply from the byproducts of full information maximum likelihood estimation without heavy computational burden. We provide practical information on the interpretation and behavior of these diagnostics for applied researchers and provide software code to facilitate their use. Simulated and empirical examples involving different kinds of model comparison scenarios encountered in cross-sectional, longitudinal, and multilevel research as well as involving different kinds of outcome distributions illustrate the generality of the proposed diagnostics. An awareness of how cases influence model selection results is shown to aid researchers in understanding how representative their sample level results are at the case level.

Intranasal oxytocin versus placebo in the treatment of adults with autism spectrum disorders: a randomized controlled trial

BackgroundThere are no effective medications for the treatment of social cognition/function deficits in autism spectrum disorder (ASD), and adult intervention literature in this area is sparse. Emerging data from animal models and genetic association studies as well as early, single-dose intervention studies suggest that the oxytocin system may be a potential therapeutic target for social cognition/function deficits in ASD. The primary aim of this study was to examine the safety/therapeutic effects of intranasal oxytocin versus placebo in adults with ASD, with respect to the two core symptom domains of social cognition/functioning and repetitive behaviors.MethodsThis was a pilot, randomized, double-blind, placebo-controlled, parallel design trial of intranasal oxytocin versus placebo in 19 adults with ASD (16 males; 33.20 ± 13.29 years). Subjects were randomized to 24 IU intranasal oxytocin or placebo in the morning and afternoon for 6 weeks. Measures of social function/cognition (the Diagnostic Analysis of Nonverbal Accuracy) and repetitive behaviors (Repetitive Behavior Scale Revised) were administered. Secondary measures included the Social Responsiveness Scale, Reading-the-Mind-in-the-Eyes Test and the Yale Brown Obsessive Compulsive Scale – compulsion subscale and quality of life (World Health Organization Quality of Life Questionnaire – emotional/social subscales). Full-information maximum-likelihood parameter estimates were obtained and tested using mixed-effects regression analyses.ResultsAlthough no significant changes were detected in the primary outcome measures after correcting for baseline differences, results suggested improvements after 6 weeks in measures of social cognition (Reading-the-Mind-in-the-Eyes Test, p = 0.002, d = 1.2), and quality of life (World Health Organization Quality of Life Questionnaire – emotion, p = 0.031, d = 0.84), both secondary measures. Oxytocin was well tolerated and no serious adverse effects were reported.ConclusionsThis pilot study suggests that there is therapeutic potential to daily administration of intranasal oxytocin in adults with ASD and that larger and longer studies are warranted.Trial registrationNCT00490802

A simultaneous equations model of finance and growth: FIML estimates for India

In the relationship between economic growth and financial development, it is generally conceded that both variables are likely to be interdependent. However, no attempt has been made so far to estimate a simultaneous equations model to test whether finance causes growth or vice versa. This article uses the Full Information Maximum Likelihood (FIML) method to estimate a two equations model of growth and finance for India to determine the strength of this interdependence. Our results show that Financial Developments (FD) have a small but significant permanent growth effect. However, there is no evidence to support the view that ‘where enterprise leads, finance follows’.

Generalized full-information item bifactor analysis.

Full-information item bifactor analysis is an important statistical method in psychological and educational measurement. Current methods are limited to single-group analysis and inflexible in the types of item response models supported. We propose a flexible multiple-group item bifactor analysis framework that supports a variety of multidimensional item response theory models for an arbitrary mixing of dichotomous, ordinal, and nominal items. The extended item bifactor model also enables the estimation of latent variable means and variances when data from more than 1 group are present. Generalized user-defined parameter restrictions are permitted within or across groups. We derive an efficient full-information maximum marginal likelihood estimator. Our estimation method achieves substantial computational savings by extending Gibbons and Hedeker's (1992) bifactor dimension reduction method so that the optimization of the marginal log-likelihood requires only 2-dimensional integration regardless of the dimensionality of the latent variables. We use simulation studies to demonstrate the flexibility and accuracy of the proposed methods. We apply the model to study cross-country differences, including differential item functioning, using data from a large international education survey on mathematics literacy.

Examining Familiarity through the Temperament and Character Inventory: A Structural Equation Modeling Analysis

The purpose of this study was to examine whether several aspects of personality, as measured by the Temperament and Character Inventory (i.e., novelty seeking, harm avoidance, reward dependency, persistence, self-directedness, cooperation, and self-transcendence), were determined in part by family factors (i.e., genes and/or familial environments). Analyses were conducted on a sample of 358 Japanese families (1224 individuals) using structural equation modeling (SEM). In our analysis, we operationally assumed that families with fewer than four children had missing data as a means of integrating different patterns of families simultaneously using full information maximum-likelihood (FIML) estimates. Results showed the influences of parent-offspring relationships, spouse and sibling correlations, gender differences, and heritability. Both familial environment and genetic sources of resemblance were suggested for temperament and character. Finally, future tasks were discussed.