Within-individual Correlations Research Articles

Detailed analysis of binges in obese women with binge eating disorder: Comparisons using multiple methods of data collection

This study compares multiple methods of assessing food intake in obese women with binge eating disorder (BED). Twelve women meeting BED criteria completed six random 24-hour dietary recalls, engaged in a laboratory binge eating episode, and completed the EDE interview. There was not a significant difference in total or macronutrient intake when binge eating episodes were assessed via the recall and laboratory methods. However, within-individual correlations were low for the size of different binge eating episodes collected by the two methods. Significantly more calories were consumed during objective than during subjective binges, and significant differences in macronutrient composition were observed. Meal patterning data collected by the EDE and the recalls were comparable. The findings suggest only moderate agreement between the methods that were examined. Future investigations with larger sample sizes are needed to examine the relationship among these different methods of assessing food intake.

How Many Patients? How Many Limbs? Analysis of Patients or Limbs in the Orthopaedic Literature: A Systematic Review

Clinical studies assessing orthopaedic interventions often include data from two limbs or multiple joints within single individuals. Without appropriate design or statistical approaches to address within-individual correlations, this practice may contribute to false precision and possible bias in estimates of treatment effect. We conducted a systematic review of the orthopaedic literature to determine the frequency of inappropriate inclusion of nonindependent limb or joint observations in clinical studies. We identified seven orthopaedic journals with high Science Citation Index impact factors and retrieved all clinical studies for 2003 for any intervention on any limb or joint. We identified 288 clinical studies, 143 of which involved two limbs or multiple joint observations from single individuals. These studies included nineteen randomized clinical trials (13%) fifty-eight two-group cohort studies (41%), and sixty-six one-group cohort studies (46%). Seventy-six (53%) of the 143 studies involved statistical comparisons between patient groups with use of tests of association, and an additional sixty studies (42%) presented estimates of proportions without statistical comparisons. Only sixteen of the seventy-six studies involving statistical comparisons involved the use of any technique or methodological approach to account for multiple, nonindependent observations. A median of approximately 13% of the patients in these studies contributed more than one observation. The median proportion of nonindependent observations to total observations (the unit of analysis) was approximately 23%. Our findings suggest that a high proportion (42%) of clinical studies in high-impact-factor orthopaedic journals involve the inappropriate use of multiple observations from single individuals, potentially biasing results. Orthopaedic researchers should attend to this issue when reporting results.

Accuracy of Single- and Multiple-Trait REML Evaluation of Data Including Non-Random Missing Records

AbstractWe examined the accuracy of single- and multiple-trait REML procedures by studying estimates of within-individual genetic correlations between an ordered categorical trait and a continuous trait. The traits were derived from simulated bivariate, normally distributed data including selectively deleted records. Ten thousand data sets were generated for each partially factorial combination of two levels of genetic correlation (0.3 and 0.6), and environmental correlation (0.3 and 0.6), and three levels of narrow-sense individual heritability (0.05, 0.15 and 0.25) and mortality (0, 10, 30 and 50%). All data sets consisted of data on 200 unrelated parents, each with 20 halfsib progenies. The accuracy of the evaluations was illustrated in terms of average bias and variation of derived correlation estimates. The average bias values generated by multiple-trait REML were generally low. In contrast, single-trait REML was sensitive to selective deletion of records and systematically underestimated the genetic correlations. For both methods, especially at low heritabilities, the magnitude of the variation was generally high, showing that there is a substantial probability of obtaining seriously misleading genetic correlation estimates if the analysis is based on a single experiment and data include non-random missing records.

Different effects of intensity and duration of locomotor activity on circadian period.

An outstanding unresolved issue in chronobiology is how the level of locomotor activity influences length of the free-running, endogenous circadian period (tau). To address this issue, the authors studied a novel model, 4 replicate lines of laboratory house mice (Mus domesticus) that had been selectively bred for high wheel-running activity (S) and their 4 unselected control (C) lines. Previous work indicates that S mice run approximately twice as many revolutions/day and exhibit an altered dopaminergic function as compared with C mice. The authors report that S mice have a tau shorter by about 0.5 h as compared with C mice. The difference in tau was significant both under constant light (control lines: tau = 25.5 h; selected: tau = 24.9 h) and under constant dark (control lines: 23.7 h; selected: 23.4 h). Moreover, the difference remained statistically significant even when the effects of running speed and time spent running were controlled in ANCOVA. Thus, something more fundamental than just intensity or duration of wheel-running activity per se must underlie the difference in tau between the S and C lines. However, despite significant difference in total wheel-running activity between females and males, tau did not differ between the sexes. Similarly, among individuals within lines, tau was not correlated with wheel-running activity measured as total revolutions per day. Instead, tau tended to decrease with average running speed but increase with time spent running. Finally, within individuals, an increase in time spent running resulted in decreased tau in the next few days, but changes in running speed had no statistically significant effect. The distinctions between effects of duration versus intensity of an activity, as well as between the among- versus within-individual correlations, are critical to understanding the relation between locomotor activity and pace of the circadian clock.

Family study of the relationship between height and cardiovascular risk factors in the STANISLAS cohort.

Short stature is a risk factor for coronary heart disease and is associated with an adverse cardiovascular profile. Mechanisms responsible for this association remain unknown. A genetic contribution to this association would imply a familial clustering between height and cardiovascular risk factors. This study investigated whether lipids and blood pressure (BP) levels shared a common familial component with height. The sample included 865 nuclear families from the French STANISLAS cohort volunteering for a free health examination between 1993 and 1994. Within-individual correlations and familial intra-trait and cross-trait correlations were estimated using the Estimating Equation technique extended to a bivariate phenotype. Height negatively correlated to total and low density lipoprotein cholesterol (LDL-C) and triglycerides in both parents and offspring, and positively correlated to high density lipoprotein cholesterol (HDL-C) in parents only. In offspring, the correlation between height and HDL-C markedly increased with sexual maturation to reach after puberty the same value as in parents. The correlation of height with systolic BP was negative in fathers and positive in sons, whereas it was non-significant in mothers and daughters. The pattern of cross-trait familial correlations between height and LDL-C was compatible with the existence of a weak transmissible component explaining the relationship between these two traits. By contrast, the pattern observed for HDL-C and triglycerides was rather compatible with the influence of shared environmental factors. No familial clustering between height and BP levels was detected. The association between short stature and increased LDL-C might be partly of familial origin.

Are within-individual causes of delinquency the same as between-individual causes?

Previous studies of the causes of delinquency have been based on between-individual correlations. This paper aims to study the causes of delinquency by comparing within-individual and between-individual correlations of risk factors with delinquency. A total of 506 boys in the oldest sample of the Pittsburgh Youth Study were followed up in seven data waves between ages 13.8 and 17.8 on average. Poor parental supervision, low parental reinforcement and low involvement of the boy in family activities were the most important causes of delinquency according to forward-lagged within-individual correlations. Poor housing was positively related to delinquency for boys living in bad neighbourhoods but not for boys living in good neighbourhoods. Forward-lagged within-individual correlations provide more valid information about the causes of delinquency than do between-individual correlations. Peer delinquency was the strongest correlate of delinquency according to between-individual correlations but was not a cause of delinquency according to forward-lagged within-individual correlations.

Life events and depression in a community sample of siblings.

The overall aim of the GENESiS project is to identify quantitative trait loci (QTLs) for anxiety/depression, and to examine the interaction between these loci and psychosocial adversity. Here we present life-events data with the aim of clarifying: (i) the aetiology of life events as inferred from sibling correlations; (ii) the relationship between life events and measures of anxiety and depression, as well as neuroticism; and (iii) the interaction between life events and neuroticism on anxiety/depression indices. We assessed the occurrence of one network and three personal life-event categories and multiple indices of anxiety/depression including General Health Questionnaire, Anhedonic Depression, Anxious Arousal and Neuroticism in a large community-based sample of2150 sib pairs, 410 trios and 81 quads. Liability threshold models and raw ordinal maximum likelihood were used to estimate within-individual and between-sibling correlations of life events. The relationship between life events and indices of emotional states and personality were assessed by multiple linear regression and canonical correlations. Life events showed sibling correlations of 0-37 for network events and between 0-10 and 0.19 for personal events. Adverse life events were related to anxiety and depression and, to a less extent, neuroticism. Trait-vulnerability (as indexed by co-sib's neuroticism, anxiety and depression) accounted for 11% and life events for 3% of the variance in emotional states. There were no interaction effects. Life events show moderate familiality and are significantly related to symptoms of anxiety and depression in the community. Appropriate modelling of life events in linkage and association analyses should help to identify QTLs for depression and anxiety.

Neural grafting reverses prenatal drug-induced alterations in hippocampal PKC and related behavioral deficits

Administration of heroin or phenobarbital to pregnant mice evokes neurochemical and behavioral deficits consequent to disruption of septohippocampal cholinergic innervation. The present study evaluates the relationship between the drug-induced biochemical changes and the behavioral deficits, applying two different approaches: neural grafting and within-individual correlations of biochemistry and behavior. Mice were exposed transplacentally to phenobarbital or heroin on gestational days 9–18 and tested in adulthood. Drug-exposed mice displayed impaired radial arm maze performance, increases in presynaptic choline transporter sites (monitored with [ 3H]hemicholinium-3 binding), upregulation of membrane-associated protein kinase C (PKC) activity, and desensitization of the PKC response to a cholinergic agonist, carbachol. Grafting of cholinergic cells to the impaired hippocampus reversed the behavioral deficits nearly completely and restored basal PKC activity and the PKC response to carbachol to normal; the drug effects on hemicholinium-3 binding were also slightly obtunded by neural grafting, but nevertheless remained significantly elevated. There were significant correlations between the performance in the eight-arm maze and both basal PKC activity and PKC desensitization, and to a lesser extent, between behavioral performance and hemicholinium-3 binding. Taken together, these findings indicate an inextricable link between the biochemical effects of prenatal drug exposure on the PKC signaling cascade and adverse behavioral outcomes. The ability of neural grafting to reverse both the drug-induced changes in PKC and behaviors linked to septohippocampal cholinergic function suggest a mechanistic link between this signaling pathway and neurobehavioral teratology caused by heroin or phenobarbital.

Analysis of repeated hemoglobin measures in full-term, normal birth weight Kenyan children between birth and four years of age. III. The Asemobo Bay Cohort Project.

Anemia is an important public health problem. During very early childhood numerous factors affect hemoglobin (Hb) concentration over time, making single cross-sectional measurements difficult to interpret when studying the natural history of anemia or evaluating anemia control strategies. We analyzed repeated Hb measures contributed by 942 Kenyan children between birth and 48 months of life using a mixed effects model, with a regression spline used to describe the population mean Hb profile, and random intercepts and slopes and first-order autoregressive correlation structure to accommodate the within-individual correlation among the repeated Hb measures. The approach facilitates the study of time-stationary and time-varying covariates that influence Hb in early life. The fitted mean Hb profile obtained from the analytic model is consistent with the observed mean Hb of the study population. Village of residence was associated with greatest difference in mean Hb at time of birth (16 versus 19 g/dL; P < 0.0001). Monthly weight-for-age was also associated with mean Hb after 3 months of age. This is the first description of an analysis strategy specifically for repeated Hb measures collected in a longitudinal field study in Africa. The strategy will facilitate improved study of time-varying covariates thought to influence pediatric anemia.

Tolerance assessment of Cistus ladanifer to serpentine soils by developmental stability analysis

Developmental instability is the result of random environmental perturbations during development. Its absence (developmental stability) depends on an organism's ability to buffer environmental disturbances. Both genotype and environment influence the phenotypic expression of developmental instability and it is susceptible to selection pressure. We studied developmental instability (as indicated by increased within-individual asymmetry of repeated traits) in vegetative and reproductive structures of three populations of Cistus ladanifer L. living in different soil substrates (serpentine, siliceous and contact zone) to detect tolerance to serpentine soils. Serpentine soils, characterized by high concentrations of heavy metals (Ni, Cr, and Co), low levels of Ca/Mg ratio and high water deficit, can adversely affect plant performance. In this study we demonstrated that asymmetry and within-plant variance were higher in the contact zone population than either the silica or serpentine populations, proving the adaptation of C. ladanifer to serpentine soils. Within-population estimates of developmental instability were concordant for both vegetative and reproductive traits. There was little or no within-individual correlation among estimates of developmental instability based on different structures, i.e., plants that had highly asymmetric leaves always had high developmental instability in translational symmetry. Radial asymmetry of petals was negatively correlated with petal size, especially in silica soil plants, providing evidence of selection for symmetric and large petals. While leaf size was positively correlated with absolute fluctuating asymmetry, suggesting selection for small or intermediate size leaves. Serpentine soils presented the largest foliar and floral traits, as well as shoot elongation, while silica soil plants had the smallest scores. On the contrary, aboveground plant biomass was larger in silica soil plants, while the contact zone plants had the lowest biomass.

Effect of Omitted Confounders on the Analysis of Correlated Binary Data

Marginal analysis using the generalized estimating equation approach is widely applied to correlated observations, as occur in studies with clusters and in longitudinal follow-up of individuals. In this article, we investigate the effect of confounding in such models. We assume that a risk factor x and a confounder z are related by a generalized linear model to the outcome y, which can be binary or ordinal. In order to investigate confounding arising from the omission of z, a joint structure for x and z must be specified. Modeling normally distributed (x,z) as sums of between- and within-individual (or cluster) components allows us to incorporate different degrees of between- and within-individual correlation. Such a structure includes, as special cases, cohort and period effects in longitudinal settings and random intercept models. The latter situation corresponds to allowing z to vary only on the between-individual (or cluster) level and to be uncorrelated with x, and leads to attenuation of the coefficient of x in marginal models with the logit and probit links. More complex situations occur when z is allowed to also vary on the within-individual (or cluster) level and when z is correlated with x. We examine the model specification and the expected bias when fitting a marginal model in the presence of the omitted confounder z. We derive general formulas and interpret the parameters and results in an ongoing cohort study. Testing for omitted covariates is also discussed.

Parametric models to assess tracking of categorical high-risk sexual behaviour for HIV infection.

Prevalence and tracking (within-individual correlation) of high-risk behaviour measured in longitudinal studies provide a complementary description of how behaviour is maintained over time. Without tracking, when population prevalences are approximately constant over time, the steadiness of the aggregate pattern is likely to mask the dynamic changes of the response occurring on an individual level. In this paper, a parametric model is proposed for the estimation of these features when the observed response is a time-stationary categorical measurement. The model extends the standard Dirichlet-multinomial model in the spirit of Prentice (1986, JASA) by allowing both negative and positive correlation among repeated categorical measurements. In addition, both the marginal category probabilities and the tracking can be modelled as a function of individual-level covariates. Details regarding likelihood based estimation and inference are provided and illustrated with data from the Multicenter AIDS Cohort Study (MACS).

Longitudinal relation between smoking and white blood cells.

Higher white blood cell counts in smokers compared with nonsmokers have been well documented, but the longitudinal relation between changes in smoking and changes in white blood cells has not been well described. Since 1984, data have been collected semiannually by the Multicenter AIDS Cohort Study (MACS), a four-center prospective cohort study of acquired immunodeficiency syndrome (AIDS) in homosexual men. The study population includes 2,435 participants who were human immunodeficiency virus (HIV) seronegative as of September 1994 and who contributed 20,918 person-visits for this analysis. For individuals who modified their smoking behavior, changes in white blood cell counts occurred primarily during the first 6 months following changes in the amount of cigarettes smoked. Among former smokers who resumed smoking, the extent of the increase in white blood cell count depended on the number of cigarettes smoked. Specifically, increases of 241, 340, and 740 cells/microliter were observed for smokers who resumed smoking < 1, 1 to < 2, and > or = 2 packs/day, respectively. Conversely, smokers who quit smoking had a decrease of white blood cell count: -32, -629, and -1,122 cells/microliter for men who previously smoked < 1, 1 to < 2, and > or = 2 packs/day, respectively. Long-term ex-smokers, however, still had higher white blood cell counts than did never smokers. There was a high within-individual correlation of white blood cell count in persons who reported a consistent level of smoking (i.e., average correlations between two white blood cell counts 6 years apart were 0.51 for never smokers, 0.48 for ex-smokers, 0.56 for men who smoked < 1 pack/day, and 0.43 for men who smoked > or = 1 pack/day). These analyses indicate an acute effect of changes in smoking on changes in white blood cell count, a residual effect of having been a smoker, and high long-term tracking for white blood cell count.

Cognition-correlation indices of gender schema: tests of validity

Four-hundred and seventy-seven subjects evaluated a set of traits and behaviors in terms of how masculine and feminine they were and in terms of how well they represented their real and ideal self-images. Within-individual correlation coefficients between these evaluations were proposed as measures of psychological gender schemata, because they would represent the degree of matching between the subjects' gender-image and ideal/real self-images of gender-related attributes. The present study aims at examining the construct validity of these measures, by testing them to psychological variables that are known to reflect gender identity. The individual difference variables used as criteria were (a) satisfaction with one's own sex, (b) general happiness, (c) self-esteem (d) gender-conflict, and (e) school and occupational achievement need. Correlations between the gender-schema indices and the criteria variables supported the construct validity of those measures. Advantages of the present measurement over the conventional simple trait approach, such as BSRI, or PAQ are discussed.

Testing the Effect of Treatment in Experiments with Correlated Binary Outcomes

This paper considers the problem of testing for treatment effect in a randomized experiment with correlated binary outcomes, representing success or failure for different "parts" of a randomized unit. Attention is restricted to tests that are based on a summary score for each individual randomized, and thus are valid regardless of the precise nature of the correlation among parts. The focus is on the efficiency of such tests under various correlation structures, with special emphasis on the case in which the correlation among parts within an individual differs across treatment groups. A class of summary score statistics is defined, and optimal testing is discussed for some simple situations. Three potential general-purpose tests also are described: (1) the ratio estimate test discussed by Henderson et al. (1988, Controlled Clinical Trials 9, 189-205); (2) a modified ratio estimate test with adjusted weighting based on the within-individual correlation between parts; (3) a test defined by applying the Mantel-Haenszel procedure to the proportion of individuals with at least one failure, stratifying by the number of parts. For these general-purpose tests, numerical calculations of asymptotic efficiency are presented under a wide range of designs and correlation structures. On the basis of these results, some practical recommendations for choosing a test are made.

Changes in T-Lymphocyte Subsets in Intravenous Drug Users With HIV-1 Infection

To evaluate changes in T-cell subsets in prevalent human immunodeficiency virus type 1 (HIV-1) seronegative and seropositive intravenous drug users (IVDUs) and in HIV-1 seropositive IVDUs with known time of seroconversion. Cohort study with a median 18-month follow-up. Community-based clinic established to study the natural history of HIV infection in IVDUs. Eight hundred fifty-nine self-referred IVDUs aged 18 through 49 years who injected drugs within the last 10 years and who did not have an AIDS (acquired immunodeficiency syndrome)--defining illness; 152 were seronegative for HIV-1, 621 were seropositive, and 86 seroconverted during the study. Proportions and absolute numbers of lymphocytes and CD3, CD4, and CD8 T cells as determined at 6-month intervals by flow cytometry and complete blood cell counts with automated differential. Median numbers of CD4 lymphocytes at enrollment were 1061/microL (1.06 x 10(9)/L) for seronegative IVDUs, 508/microL for seropositive IVDUs, and 733/microL for those who seroconverted (enrolled a median of 4.5 months after seroconversion); the corresponding figures for CD8 lymphocytes were 628, 894, and 889/microL, respectively. Median rates of decline in absolute numbers and percentages of CD4 lymphocytes per 6 months were 7.6/microL (0.0%) for seropositive IVDUs and 55.1/microL (1.9%) for IVDUs who seroconverted (median follow-up after seroconversion was 12 months). Multivariate regression analysis that incorporated the within-individual correlation of the CD4 lymphocyte counts showed no significant change in these cells over time and no change due to use of drugs. Our data suggest that progression of HIV-1 infection in IVDUs, as reflected in decline of CD4 cell counts, is no more rapid than that reported for other risk groups.

Analysis of Longitudinal Data with Unmeasured Confounders

Confounding in longitudinal or clustered data creates special problems and opportunities because the relationship between the confounder and covariate of interest may differ across and within individuals or clusters. A well-known example of such confounding in longitudinal data is the presence of cohort and period effects in models of aging in epidemiologic research. We first formulate a data-generating model with confounding and derive the distribution of the response variable unconditional on the confounder. We then examine the properties of the regression coefficient for some analytic approaches when the confounder is omitted from the fitted model. The expected value of the regression coefficient differs in across- and within-individual regression. In the multivariate case, within- and between-individual information is combined and weighted according to the assumed covariance structure. We assume compound symmetry in the fitted covariance matrix and derive the variance, bias, and mean squared error of the slope estimate as a function of the fitted within-individual correlation. We find that even in this simplest multivariate case, the trade-off between bias and variance depends on a large number of parameters. It is generally preferable to fit correlations somewhat above the true correlation to minimize the effect of between-individual confounders or cohort effects. Period effects can lead to situations where it is advantageous to fit correlations that are below the true correlation. The results highlight the trade-offs inherent in the choice of method for analysis of longitudinal data, and show that an appropriate choice can be made only after determining whether within- or between-individual confounding is the major concern.

Development of the motor memory trace and control program.

A theoretical explanation of the development of the memory trace constrained to fit into a computer analog requires a coded reference "image" of the intended motor act, a tentative control program (memory trace) based on mobilizing and coordinating various pre-established motor-control subroutines, and an afferent neural description of the act as actually performed. The afferent information is non-consciously compared with the reference image, and the resulting discrepancy used to modify the control program so as to reproduce the reference image more exactly on the next attempt. The discrepancy consists of the sum of two types of differences, between image and program (present during learning) and the error difference between program and performance. The trial-by-trial discrepancies cumulate in the memory trace, so that during both learning and postlearning practice the error component results in a within-individual odd-even trial correlation. This constitutes a mechanism for a remoteness effect in the inter-trial correlation matrix generated in a group of individuals during post-learning as well as learning conditions. Fifty-one individuals performed 50 trials on a simple arm movement task; 35 of these (the learners) showed individual systematic improvement during the first 20 trials but no improvement during the last 24; 16 (the non-learners) failed to improve with practice. Separate correlational analyses (both within- and inter-individual) of the two groups resulted in a variety of patterns that were predicted a priori in the learners by the theory, as did the 49 learners of another sample of 51 who performed a more complicated arm movement task. Also, as predicted, the non-learners failed to produce organized correlation patterns.

Geographic Variation Analysis of Dipodomys ordii Using Nonmetric Cranial Traits

Geographic variation in 18 nonmetric cranial traits was assessed for 11 populations of Dipodomys ordii from Oklahoma, Texas, and New Mexico. Frequencies of all traits were examined for sex, size, and side dependencies in a geographically restricted sample of 272 crania. Two traits were significantly affected by sex, two by cranial size, and two exhibited different frequencies on the left side of the skull than on the right. Relatively few within-individual and among-population intertrait correlations were found. The distance measure used in testing for interpopulation trait frequency differences was the Mean Measure of Divergence (MMD), as modified by Green and Suchey (1976). Nonmetric multidimensional scalings of distance matrices derived from each of four trait subsets and a principal components analysis based on one subset gave similar results. The 11 samples were displayed in three-dimensional summaries approximately as expected on the basis of their geographic distribution. Because nonmetric traits may be adaptively neutral, frequency differences between populations may result from genetic drift. Nonmetric traits are clearly of value in studies of geographic variation in Dipodomys . I follow Sjovold (1977) in suggesting that consideration be given to each trait with regard to how strongly it is influenced by dependencies (e.g., sex and size) that may confound interpretation of results.

Ovarian arterial blood velocity measured with Doppler ultrasonic transducers in conscious ewes.

Changes in the mean velocity of ovarian arterial blood flow during the oestrous cycle were monitored in 5 ewes by the use of doppler ultrasonic transducers chronically implanted around the ovarian arteries. In arteries supplying ovaries with a corpus luteum (CL), the velocity was minimal from Day -1 to Day 2 inclusive (Day 0 = day of oestrus), increased steadily until Day 13 (-4) and then declined precipitously. In contralateral arteries supplying ovaries without a CL, the velocity remained at a consistently low level throughout the oestrous cycle. In each ewe, arterial blood velocity to the ovary with a CL and progesterone levels in peripheral plasma were highly correlated; within-individual correlations falling between r = 0.830 (P < 0.001, n = 15) and r = 0.936 (P < 0.001, n = 15). The changes in the velocity of the arterial supply to the ovulatory ovary and the pattern to that for weight of luteal tissue in similar ewes.