Mijke Rhemtulla

Planned Missing Data Designs in Educational Psychology Research

Although missing data are often viewed as a challenge for applied researchers, in fact missing data can be highly beneficial. Specifically, when the amount of missing data on specific variables is carefully controlled, a balance can be struck between statistical power and research costs. This article presents the issue of planned missing data by discussing specific designs (i.e., multiform designs, longitudinal wave-missing designs, and 2-method measurement designs), introducing the power and cost benefits of such scenarios to applied education and educational psychology researchers.

Genomic SEM Provides Insights into the Multivariate Genetic Architecture of Complex Traits

Methods for using GWAS to estimate genetic correlations between pairwise combinations of traits have produced “atlases” of genetic architecture. Genetic atlases reveal pervasive pleiotropy, and genome-wide significant loci are often shared across different phenotypes. We introduce genomic structural equation modeling (Genomic SEM), a multivariate method for analyzing the joint genetic architectures of complex traits. Using formal methods for modeling covariance structure, Genomic SEM synthesizes genetic correlations and SNP-heritabilities inferred from GWAS summary statistics of individual traits from samples with varying and unknown degrees of overlap. Genomic SEM can be used to identify variants with effects on general dimensions of cross-trait liability, boost power for discovery, and calculate more predictive polygenic scores. Finally, Genomic SEM can be used to identify loci that cause divergence between traits, aiding the search for what uniquely differentiates highly correlated phenotypes. We demonstrate several applications of Genomic SEM, including a joint analysis of GWAS summary statistics from five genetically correlated psychiatric traits. We identify 27 independent SNPs not previously identified in the univariate GWASs, 5 of which have been reported in other published GWASs of the included traits. Polygenic scores derived from Genomic SEM consistently outperform polygenic scores derived from GWASs of the individual traits. Genomic SEM is flexible, open ended, and allows for continuous innovations in how multivariate genetic architecture is modeled.

Fatherhood alters gene expression within the MPOA

Female parenting is obligate in mammals, but fathering behavior among mammals is rare. Only 3–5% of mammalian species exhibit biparental care, including humans, and mechanisms of fathering behavior remain sparsely studied. However, in species where it does exist, paternal care is often crucial to the survivorship of offspring. The present study is the first to identify new gene targets linked to the experience of fathering behavior in a biparental species using RNA sequencing. In order to determine the pattern of gene expression within the medial preoptic area that is specifically associated with fathering behavior, we identified differentially expressed genes in male prairie voles (Microtus ochrogaster) that experienced one of three social conditions: virgin males, pair bonded males, and males with fathering experience. Differentially expressed genes from each comparison (i.e., Virgin vs Paired, Virgin vs Fathers, and Paired vs Fathers) were evaluated using the Gene Ontology enrichment analysis, and Kegg pathways analysis to reveal metabolic pathways associated with specific differentially expressed genes. Using these tools, we identified a group of genes that are differentially expressed in voles with different amounts of social experience. These genes are involved in a variety of processes, with particular enrichment in genes associated with immune function, metabolism, synaptic plasticity, and the remodeling of dendritic spines. The identification of these genes and processes will lead to novel insights into the biological basis of fathering behavior.

Robustness and replicability of psychopathology networks

Network approaches to psychopathology hold that mental disorders arise from the interplay between symptoms in a network structure1, 2. In the past few years, statistical techniques that estimate networks were developed and applied to many disorders3. As empirical findings start to accumulate, the question arising is which of these findings are robust and replicable. Here we evaluate the state of psychopathological network research based on three methodological criteria: model quality, precision, and replicability.

Longitudinal Trajectories and Inter-parental Dynamics of Prairie Vole Biparental Care

For altricial mammalian species, early life social bonds are constructed principally between offspring and their mothers, and the mother-offspring relationship sets the trajectory for offspring bio-behavioral development. In the rare subset of monogamous and biparental species, offspring experience an expanded social network which includes a father. Accordingly, in biparental species fathers also have the potential to influence trajectories of offspring development. Previous semi-natural and laboratory study of one monogamous and biparental species, the prairie vole (Microtus ochrogaster), has given insight into the role that mothers and fathers play in shaping behavioral phenotypes of offspring. Of particular interest is the influence of biparental care in the development of monogamous behavior in offspring. Here, we first briefly review that influence. We then present novel research which describes how parental investment in prairie voles changes across sequential litters of pups, and the extent to which it is coordinated between mothers and fathers. We use approximately 6 years of archival data on prairie vole parenting to investigate trajectories and inter-parent dynamics in prairie vole parenting. We use a series of latent growth models to assess the stability of parental investment across the first 4 l. Our findings suggest that prairie voles display sexually dimorphic patterns of change in parental behavior: mothers’ investment declines linearly whereas fathers’ pattern of change is characterized by initial decline between litters 1 and 2 with subsequent increase from litters 2 to 4. Our findings also support a conclusion that prairie vole paternal care may be better characterized as compensatory—that is, fathers may compensate for decline in maternal investment. Opposing trends in investment between mothers and fathers ultimately imply stability in offspring investment across sequential litters. These findings, combined with previous studies, generate a hypothesis that paternal compensation could play an important role in maintaining the development of monogamous behavioral phenotypes in individual offspring and across cohorts of those offspring. Understanding longitudinal and inter-individual dynamics of complex social behaviors is critical for the informed investigation of both proximate and ultimate mechanisms that may subserve these behaviors.

How to Measure Nothing

Maul’s target article is insightful and enlightening, and it presents a number of very important recommendations for psychometric practice. While we overwhelmingly agree with the arguments made in the second half of the paper, however, we are not convinced that the results presented in the first half of the paper are really damning evidence against the adequacy of psychometric processes. The setup of Maul’s research design appears to be based on the idea that, since the word gavagai doesn’t refer to anything, the psychometric items that query participants’ ideas about gavagai will not measure anything more or less by definition. Moreover, Maul appears to think that standing psychometric practices should reveal this: “If ever there were a time when a theory deserved to be falsified, this would appear to be it.” From the fact that standard psychometric practices do not reveal any significant problems in the questionnaire, Maul concludes that there must be something deeply wrong with these practices. This conclusion, however, does not follow for two reasons. First, it is not clear that the gavagai questionnaire measures nothing, and hence, it is not obvious that the premise underlying Maul’s argumentation is fulfilled. Second, most psychometric practices are based on the antecedent assumption that researchers are able to target a given attribute with a set of items (i.e., most psychometric practice assumes that test constructors have at least to some extent built validity into the test through item formulation and selection); they are not, however, designed to expose the falsity of that assumption. We therefore think it is useful to consider more deeply what these results reveal about response processes and validity. Maul interprets his results as showing that it is so easy to get a set of well-behaving items that it can be done even without item content. In fact, he suggests that “favorable-looking results of covariance-based statistical procedures (such as high-reliability estimates and fit to unidimensional latent variable models) should be regarded more as a default expectation for survey response data than as positive evidence for the validity of an instrument as a measure of a psychological attribute.” But experience tells us that’s not right: Researchers typically have to try hard (pilot testing, cutting items, rewording items, etc.) to get item sets that behave well, and many scales that are proposed in the literature turn out to be best described by a multidimensional factor model. It is not at all a given that any random set of items subjected to a factor analysis will result in a well-fitting unidimensional model with high factor loadings. So there are actually quite important questions here that Maul scarcely addresses—namely, Why do these nonsense items behave so well? Why are the responses so structured? What response process was tapped by the seemingly nonsensical items? There is a wealth of research showing that respondents in psychological research are willing to work with the investigator and behave according to the demand characteristics of the experimental design. What were the Mechanical Turk workers thinking as they filled out these nonsense items? What did they hypothesize that Andrew Maul wanted to know about them? We do not know the answers to these questions, and as a result it is far from easy to determine what Maul’s results really entail. It may be the case that the results show that standing psychometric practices are deeply deficient. For instance, participants might offer a truly random answer to the first question and stick with it for the following nine questions out of a desire to act consistently. This strategy is consistent with Maul’s interpretation that “at least in the context of responding to

Normal Theory GLS Estimator for Missing Data: An Application to Item-Level Missing Data and a Comparison to Two-Stage ML

Structural equation models (SEMs) can be estimated using a variety of methods. For complete normally distributed data, two asymptotically efficient estimation methods exist: maximum likelihood (ML) and generalized least squares (GLS). With incomplete normally distributed data, an extension of ML called “full information” ML (FIML), is often the estimation method of choice. An extension of GLS to incomplete normally distributed data has never been developed or studied. In this article we define the “full information” GLS estimator for incomplete normally distributed data (FIGLS). We also identify and study an important application of the new GLS approach. In many modeling contexts, the variables in the SEM are linear composites (e.g., sums or averages) of the raw items. For instance, SEMs often use parcels (sums of raw items) as indicators of latent factors. If data are missing at the item level, but the model is at the composite level, FIML is not possible. In this situation, FIGLS may be the only asymptotically efficient estimator available. Results of a simulation study comparing the new FIGLS estimator to the best available analytic alternative, two-stage ML, with item-level missing data are presented.