Cross-lagged Panel Correlation Research Articles

Introducing Continuous Time Meta-Analysis (CoTiMA)

Meta-analysis of panel data is uniquely suited to uncovering phenomena that develop over time, but extant approaches are limited. There is no straightforward means of aggregating findings of primary panel studies that use different time lags and different numbers of waves. We introduce continuous time meta-analysis (CoTiMA) as a parameter-based approach to meta-analysis of cross-lagged panel correlation matrices. CoTiMA enables aggregation of studies using two or more waves even if there are varying time lags within and between studies. CoTiMA thus provides meta-analytic estimates of cross-lagged effects for a given time lag regardless of the frequency with which that time lag is used in primary studies. We describe the continuous time underpinnings of CoTiMA, its advantages over discrete-time, correlation-based meta-analysis of structural equation models (MASEM), and how CoTiMA would be applied to meta-analysis of panel studies. An example is then used to illustrate the approach. We also conducted Monte Carlo simulations demonstrating that bias is larger for time category–based MASEM than for CoTiMA under various conditions. Finally, we discuss data requirements, open questions, and possible future extensions.

Psychotic experiences and mood episodes predict each other bidirectionally: a 6-year follow-up study in a community-based population.

Psychotic experiences (PEs) are not exclusive to psychotic disorders and highly correlated with mood episodes. In this representative general population-based study, longitudinal bidirectional associations between the extended psychosis phenotype and mood episodes were investigated, accounting for other possible causes. Households were contacted in a multistage clustered probability sampling frame covering 11 districts and 302 neighbourhoods at baseline (n = 4011) and at 6-year follow-up (n = 2185). Participants were interviewed with the relevant sections of the composite international diagnostic interview both at baseline and at follow-up. Sociodemographic, familial and environmental risk factors associated with the extended psychosis phenotype and mood episodes were assessed. Logistic regression and cross-lagged panel correlation models were used for the associations between the extended psychosis phenotype and mood episodes. PEs were associated with subsequent depressive and manic episodes. There was bidirectionality in that mood episodes were associated with subsequent PEs, and PEs were associated with subsequent mood episodes. The associations occurred in a sub-additive pattern. There were substantial synchronous and cross-lagged correlations between these psychopathology domains, with reciprocally similar cross-lagged correlations. Familial risk and adverse life events were associated with both psychopathology domains, whereas some sociodemographic risk factors and alcohol/cannabis use were associated with only one domain. The sub-additive bidirectional associations between PEs and mood episodes over time and the similarity of cross-lagged correlations are suggestive of mutually causal connections between affective and psychotic domains of psychopathology.

Longitudinal stability of social cognition in schizophrenia: A 5-year follow-up of social perception and emotion processing

BackgroundIndividuals with schizophrenia exhibit marked and disproportional impairment in social cognition, which is associated with their level of community functioning. However, it is unclear whether social cognitive impairment is stable over time, or if impairment worsens as a function of illness chronicity. Moreover, little is known about the longitudinal associations between social cognition and community functioning. MethodForty-one outpatients with schizophrenia completed tests of emotion processing (Mayer-Salovey-Caruso Emotional Intelligence Test, MSCEIT) and social perception (Relationships Across Domains, RAD) at baseline and approximately five years later. Stability of performance was assessed using paired t-tests and correlations. Longitudinal associations between social cognition and community functioning (Role Functioning Scale, RFS) were assessed using cross-lagged panel correlation analysis. ResultsPerformance on the two social cognition tasks were stable over follow-up. There were no significant mean differences between assessment points [p's≥0.20, Cohen'sd's≤|0.20|], and baseline performance was highly correlated with performance at follow-up [ρ's≥0.70, ICC≥0.83, p's<0.001]. The contemporaneous association between social cognition and community functioning was moderately large at follow-up [ρ=0.49, p=0.002]. However, baseline social cognition did not show a significant longitudinal influence on follow-up community functioning [z=0.31, p=0.76]. ConclusionsThese data support trait-like stability of selected areas of social cognition in schizophrenia. Cross-lagged correlations did not reveal a significant unidirectional influence of baseline social cognition on community functioning five years later. However, consistent with the larger literature, a moderately large cross-sectional association between social cognition and community functioning was observed. Based on stability and cross-sectional associations, these results suggest that social cognition might have short-term implications for functional outcome rather than long-term consequences.

Mathematical Creativity and Mathematical Aptitude: A Cross-Lagged Panel Analysis

Cross-lagged panel correlation (CLPC) analysis has been used to identify causal relationships between mathematical creativity and mathematical aptitude. For this study, 480 8th standard students were selected through a random cluster technique from 9 intermediate and high schools of Varanasi, India. Mathematical creativity and mathematical aptitude tests were administered twice, 4 months apart. The CLPC analysis uncovered a significant relationship between mathematical creativity and mathematical aptitude. It appears that mathematical creativity was a cause of mathematical aptitude, rather than the converse (i.e., higher mathematical creativity leading to higher mathematical aptitude). Limitations and future directions are outlined.

Is there a causal relation between mathematical creativity and mathematical problem-solving performance?

The relationship between mathematical creativity (MC) and mathematical problem-solving performance (MP) has often been studied but the causal relation between these two constructs has yet to be clearly reported. The main purpose of this study was to define the causal relationship between MC and MP. Data from a representative sample of 480 eighth-grade students were analysed using a cross-lagged panel correlation (CLPC) design. CLPC attempts to rule out plausible alternative explanation of a causal effect. The result suggests that significant predominant causal relationship was found between MC and MP. It indicates that MP was found to be a cause of MC than the converse.

Modeling emotion-creativity interaction following brief training

One form of meditation training, the integrative body-mind training (IBMT) has been shown to improve attention, reduce stress and change self-reports of mood [1]. Here, we examine whether short-term IBMT can improve performance related to creativity and determine the role that mood may play in such improvement using the cross-lagged models. Forty healthy Chinese undergraduates were randomly assigned to short-term IBMT group or a relaxation training (RT) control group. Mood and creativity performance were assessed by the Positive and Negative Affect Schedule [2] and Torrance Tests of Creative Thinking (TTCT) [3] respectively. ANOVAs revealed a group (IBMT vs. RT) × session (pre-training vs. post-training) interaction effect [F(1, 37) = 14.853; p < .01] and a session main effect [F(1, 37) = 36.156; p < .01] for TTCT. These results indicated that short-term (30 min per day for 7 days) IBMT improved creativity performance in the divergent thinking task than RT. The ANOVAs also revealed a group (IBMT vs. RT) × session (pre-training vs. post-training) interaction effect and a session main effect for positive affect (PA) and negative affect (NA) (all p < .01), indicating better emotional regulation than RT. In addition, the cross-lagged models [4] were used to explore the causal sequence between PA score and TTCT score (Figure ​(Figure1A)1A) and between NA score and TTCT score (Figure ​(Figure1B)1B) before and after IBMT training. The synchronous correlations (r PA-before × TTCT-before = .468, r PA-after × TTCT-after = .533; r NA-before × TTCT-before = -.499, r NA-after × TTCT-after = -.633) and the autocorrelations (r PA-before × PA-after = .823, r TTCT-before × TTCT-after = .591; r NA-before × NA-after = .705, r TTCT-before × TTCT-after = .591) were high in magnitude and statistically significant in the non-cross direction, which provides preliminary support for cross-lagged panel correlation. Figure 1 The causal sequence between PA score and TTCT score (A) and between NA score and TTCT score (B) before and after IBMT training. Ellipses indicate measured variables; Arrows depict hypothesized directional or “causal” links/associations; ... As predicted, PA had a positive cross-lagged impact on TTCT, which indicated a causal influence from positive mood changes to the creativity changes in IBMT group but not in RT group. In addition, NA had a negative cross-lagged impact on TTCT, which indicated a causal influence from negative mood changes to the creativity changes in the IBMT group but not in RT group.

Teacher–student relationship and academic achievement: a cross-lagged longitudinal study on three different age groups

Although the relation between teacher–student relationship and academic achievement has been addressed in many studies, the reciprocal relations between these two constructs have not been explored sufficiently. The aim of the present study was to test three competing models that hypothesized directionality of influence in relations between teacher acceptance, student-perceived teacher support, and academic achievement. Eight hundred sixteen students from 3 different grade levels in Slovenian elementary and secondary schools, covering the age range from late childhood through early-to-middle adolescence, participated in the study at the beginning and at the end of the school year. Structural equation modeling was used. Different models of relations between teacher acceptance, student-perceived teacher support, and academic achievement were analyzed using the cross-lagged panel correlation technique. The results supported the hypothesized reciprocal model, indicating the relation between teacher acceptance and academic achievement in both directions. Student-perceived teacher personal support partially mediates the relation between teacher acceptance and achievement in both directions on the whole sample but not in specific age groups. Finally, we found some age-specific differences in the relations between teacher acceptance, teacher personal support, and achievement.

Bidirectional influences between maternal and paternal parenting and children's disruptive behaviour from kindergarten to grade 2

Empirical evidence suggests that children's disruptive behaviour (CDB) and quality of parenting influence one another bidirectionally. However, few studies have considered the separate contribution of the mother–child and father–child relationships to disruptive behaviours within a longitudinal context. Against this background, the reciprocal influence between CDB and five dimensions of maternal and paternal parenting was examined from kindergarten through grade 2 in a community sample of 644 children. These relationships were investigated using cross-lagged panel correlation and reciprocal effect analysis. Differences and similarities emerged in the direction of effects linking maternal and paternal parenting and CDB. At school entry, a reciprocal parent–child influence was evident for mothers only. However, as children grew older, a unidirectional effect from CDB to parenting was noted for both fathers and mothers. Implications for future research focusing on the link between CDB and the parent–child relationship are discussed.

The effect of attorneys' work with trauma-exposed clients on PTSD symptoms, depression, and functional impairment: A cross-lagged longitudinal study.

To date, few studies have examined mental health consequences among attorneys exposed to clients' traumatic experiences. A longitudinal, 2-wave, cross-lagged study was used in a cohort of attorneys (N = 107) from the Wisconsin State Public Defender's Office. We assessed changes in posttraumatic stress disorder (PTSD), depression, and functional impairment over a 10-month period and tested the effects of intensity of contact with trauma-exposed clients on symptom levels over time. Attorneys demonstrated strong and significant symptom stability over time in PTSD, depression, functional impairment, and levels of exposure. Analyses involving cross-lagged panel correlation structural equation modeling path models revealed that attorneys' levels of exposure to trauma-exposed clients had significant positive effects, over time, on PTSD, depression, and functional impairment. Gender, age, years on the job, and office size did not predict any of the outcomes. Level of exposure to trauma-exposed clients predicted reduction of weekly working hours over time, but there was no reciprocal relationship between PTSD, depression, and functional impairment and level of exposure over time. These findings underscore the central role of exposure to trauma-exposed clients in predicting mental health outcomes and emphasize the need to support attorneys by managing the intensity of exposure as well as addressing emerging symptoms.

Relations Between Specific and General Word Learning

A one-year longitudinal study was performed to test the hypothesis that children's word-specific learning of regular words is a causal determinate in their understanding and use of simple correspondence rules in reading and spelling. Kindergarten and first-grade children were asked to read and spell real words and matched pseudowords in three sessions. Cross-lagged panel correlation analyses suggested that kindergarten children's success in learning letter and sound correspondences in real words may be influencing their learning to use letter and sound correspondences in pseudowords, from Session 1 to 2 in reading, and from Session 1 to 2 and Session 2 to 3 in spelling. After this, children's success in real words no longer seems to affect their performance on pseudowords.

A longitudinal examination of the parent–child distress relationship in children with juvenile rheumatic disease.

To examine the longitudinal relationships between parent and child distress in a sample of children with juvenile rheumatic diseases (JRDs). Cross-lagged panel correlation analysis tested the temporal precedence of parent distress versus child distress over a 1-year period. Thirty-seven children (ages 9-17 years; 22 girls) diagnosed with JRD and their parents completed self-report measures on 2 occasions (assessment interval M = 12 months). Child Depression Inventory and Brief Symptom Inventory. Significant cross-sectional parent-child distress associations were observed at both time points. Moreover, Time 1 parent distress predicted child distress at Time 2 after child-reported functional ability was controlled; Time 1 child distress was unrelated to Time 2 parent distress. Cross-lagged panel correlations demonstrated the temporal precedence of parent distress relative to child distress in the parent-child distress relationship. These preliminary findings underscore the importance of parent distress in parent-child transactional adjustment, and suggest a predominant role for parent distress in children's adjustment to JRDs.

The effects of insecure attachment orientations and perceived social support on posttraumatic stress and depressive symptoms among civilians exposed to the 2009 Israel–Gaza war: A follow-up Cross-Lagged panel design study

A follow-up Cross-Lagged-design was used to test the effects of attachment orientations and perceived social support on posttraumatic stress disorder (PTSD) and major depressive disorder symptoms (MDD) in a sample of 135 Israeli students who were evacuated from a university campus located near the Israel–Gaza border in response to increased missile-fire in the area. An internet-based data collection procedure enabled the simultaneous survey of evacuees located up to 40 km from the border at war, both during the fighting and 4 months after the ceasefire. Proximity to the border did not affect levels of PTSD or MDD symptoms, attachment orientation, or levels of perceived social support. Analyses involving Cross-Lagged Panel Correlation (CLPC) path models revealed that Attachment–Anxiety had significant positive effects on PTSD, MDD, and perceived social support. Neither PTSD nor MDD nor perceived social support had any reciprocal follow-up effect on Attachment–Anxiety. These findings underscore the central role of individual trait personality differences in predicting changes in both mental health problems and interpersonal relations over time, following exposure to trauma.

Parent–Child Relationships and ADHD Symptoms: A Longitudinal Analysis

Evidence both from psychological research and clinical intervention studies suggests that there are bidirectional influences between overt child behavior problems and parent-child relations. Very little research however, has considered the pattern of relations that exists between Attention Deficit Hyperactivity Disorder (ADHD) and the parent-child relationship within a longitudinal context. Using a longitudinal community sample from the United Kingdom which included 194 school aged children (46% male and 54% female) and both parents, this study examined the relationship between child ADHD symptoms and displays of rejection in the parent-child relationship. These relationships were investigated separately for mothers and fathers using cross-lagged panel correlation and reciprocal effects analysis. Mothers and fathers reported on ADHD symptoms and children reported on their feelings of rejection in the mother-child and father-child relationships. Results suggested differences in the direction of effects linking mother- and father-child rejection and child ADHD symptoms; with ADHD symptoms affecting the mother-child relationship and the converse pattern of effects noted for fathers. Implications for future research focusing on the link between ADHD symptoms and parent-child relationships are discussed.

Causal connections in the acquisition of an orthographic rule: a test of Uta Frith's developmental hypothesis

In a longitudinal study we tested Frith's causal hypothesis that children first gain orthographic knowledge through reading and then later, as a consequence, through spelling. Children from Years 2 and 3 were tested three times over two years on their reading and spelling of pseudo-words which conformed to the conditional orthographic rule, the 'final -e' or 'split-digraph' rule. Cross-lagged panel correlation analyses suggested that the children's success in reading split-digraph words was a causal determinant of their learning to use split-digraphs in spelling, in the 7- to 8-year period and, with one year-group but not with the other, in the 8- to 9-year period. In the 9- to 10-year period children's success in reading no longer seemed to affect their spelling. These results strongly support Frith's causal hypothesis about the development of orthographic knowledge.

Modeling Unmeasured Third Variables in Longitudinal Studies

This article discusses techniques to account for unmeasured third variables in longitudinal designs. As an extension of the synchronous common factor (SCF) model, which is the basis of the cross-lagged panel correlation technique, a series of less restrictive synchronous common factor (LRSCF) models are introduced. LRSCF models can be tested by using structural equation modeling. Although both SCF and LRSCF models rest on a 2-wave design, the SCF model requires only 2 variables measured at each wave, whereas LRSCF models require 3 variables. This enables several restrictions of the SCF model to be relaxed. Among others, a particular advantage is that the strength of lagged relations among the variables can be estimated. It is recommended that LRSCF models be applied routinely whenever possible third variables might have been omitted.

Life and Job Satisfaction: A Two-Wave Panel Study

Abstract A time-related relationship between job and life satisfaction over 13 months was examined in 299 professional employees of a British engineering company, using cross-lagged panel correlation and latent variable LISREL analysis. Although the cross-lagged correlations were significantly positive, the difference between them was not significant, implying a spurious or reciprocal relationship between the two variables. The cross-lagged path coefficients in the LISREL models were not significantly positive, suggesting that the relationship between these two variables was spurious and due to error variance. Test-retest coefficients for both variables were moderately positive, showing that the relative ranking of individuals on these variables was fairly stable.

Psychological distress and Neuroticism: A two‐wave panel study

The temporal relationship between psychological distress and Neuroticism, both measured at two points eight months apart, was examined in a representative sample of 225 adult residents in Canberra, using cross-lagged panel correlation and latent variable LISREL analysis. Psychological distress was assessed by the 30-item General Health Questionnaire (GHQ-30). Because the synchronous and autocorrelations differed significantly, the interpretation of the significant difference between the cross-lagged correlations was problematic. The cross-lagged path coefficients in the LISREL models were not significantly positive, suggesting that the temporal relationship between these two variables was spurious and due to error variance. The test-retest correlation was significantly higher for Neuroticism than for the GHQ-30, implying that Neuroticism is more a measure of a relatively stable personality characteristic and the GHQ-30 more one of transient psychological distress.

Alternative hierarchies of the attitude-behavior relationship: The impact of brand commitment and habit

The theory of reasoned action and the low-involvement hierarchy model were tested using cross-lagged panel correlation. It was proposed that the low-involvement hierarchy model would more accurately reflect the behavior of low-brand-committed individuals and that the theory of reasoned action would more accurately reflect the behavior of high-brand-committed individuals. Although the findings did not completely support these hypotheses, the general thrusts of the hypotheses appear to be supported. Additionally, behavior was hypothesized and shown to influence habit for all individuals.

Developmental Resources and Depression in the Elderly

A longitudinal study was carried out with 28 clinically depressed and 28 mentally healthy older adults, matched on age, sex, and years of education. Developmental resources and depression were measured on three occasions using the Developmental Resources of Later Adulthood and the Center for Epidemiological Studies Depression scales. A 2 X (3) repeated measures factorial analysis of variance indicated that the depressed group had significantly lower developmental scores across all three time periods. Results of cross-lagged panel correlation analyses suggested that the direction of the relationship between developmental resources and depressive symptoms differed markedly between the two groups. The mentally healthy group showed a significant causal tendency for developmental resources to influence level of depression; a reverse trend was noted in the depressed group.

Is There a Causal Relation between Self-Concept and Academic Achievement?

The relation between self-concept and achievement has often been studied, but the causal relation between these two constructs is still unclear. The purpose of this study was to define better the causal relation between self-concept and academic achievement. Longitudinal data from a large, representative sample of high school students were analyzed using a cross-lagged panel correlation (CLPC) design. CLPC attempts to rule out plausible alternative explanations of a causal effect. The results suggest that there may not be a causal relation between self-concept and achievement, but that a third variable may be causally predominant over both self-concept and achievement.