A comparative study of different covariance structure models for the analysis of repeated measurement data

In repeated measurement data, the variables are not independent, and a certain auto-correlation typically exists between different levels of repeated measurement factors. The random error is composed of at least two parts, i.e. the individual random effect and the intra-individual multi-repeated measurement effect. Traditional statistical analysis methods (such as the [Formula: see text]-test and the one-way analysis of variance) are not applicable. The linear mixed model has been widely applied for the analysis and design of repeated measurement data. This paper focuses on medical examples and describes the selection of a covariance structure for the linear mixed model of repeated measurement in the modeling of different variance–covariance structures. By selecting different covariance structures, we can perform the parameter estimation and statistical test for the fixed effect of repeated measurement data, the parameters of random effects, and the covariance matrix. The results are analyzed and compared to provide a reference for applying the linear mixed model of repeated measurement to medical research.

Download Full-text

The generalizability crisis

Behavioral and Brain Sciences ◽

10.1017/s0140525x20001685 ◽

2020 ◽

pp. 1-37

Author(s):

Tal Yarkoni

Keyword(s):

Quantitative Analysis ◽

Statistical Inference ◽

Random Effects ◽

False Positive ◽

Mixed Model ◽

Linear Mixed Model ◽

Random Effect ◽

Statistical Procedures ◽

Subject Variability ◽

Replication Crisis

Abstract Most theories and hypotheses in psychology are verbal in nature, yet their evaluation overwhelmingly relies on inferential statistical procedures. The validity of the move from qualitative to quantitative analysis depends on the verbal and statistical expressions of a hypothesis being closely aligned—that is, that the two must refer to roughly the same set of hypothetical observations. Here I argue that many applications of statistical inference in psychology fail to meet this basic condition. Focusing on the most widely used class of model in psychology—the linear mixed model—I explore the consequences of failing to statistically operationalize verbal hypotheses in a way that respects researchers' actual generalization intentions. I demonstrate that whereas the "random effect" formalism is used pervasively in psychology to model inter-subject variability, few researchers accord the same treatment to other variables they clearly intend to generalize over (e.g., stimuli, tasks, or research sites). The under-specification of random effects imposes far stronger constraints on the generalizability of results than most researchers appreciate. Ignoring these constraints can dramatically inflate false positive rates, and often leads researchers to draw sweeping verbal generalizations that lack a meaningful connection to the statistical quantities they are putatively based on. I argue that failure to take the alignment between verbal and statistical expressions seriously lies at the heart of many of psychology's ongoing problems (e.g., the replication crisis), and conclude with a discussion of several potential avenues for improvement.

Download Full-text

Response transformations for random effect and variance component models

Statistical Modelling ◽

10.1177/1471082x20966919 ◽

2020 ◽

pp. 1471082X2096691

Author(s):

Amani Almohaimeed ◽

Jochen Einbeck

Keyword(s):

Maximum Likelihood ◽

Random Effects ◽

Mixed Model ◽

Linear Mixed Model ◽

Random Effect ◽

Statistical Technique ◽

Response Distribution ◽

Level Data ◽

Variance Component Models ◽

Response Transformation

Random effect models have been popularly used as a mainstream statistical technique over several decades; and the same can be said for response transformation models such as the Box–Cox transformation. The latter aims at ensuring that the assumptions of normality and of homoscedasticity of the response distribution are fulfilled, which are essential conditions for inference based on a linear model or a linear mixed model. However, methodology for response transformation and simultaneous inclusion of random effects has been developed and implemented only scarcely, and is so far restricted to Gaussian random effects. We develop such methodology, thereby not requiring parametric assumptions on the distribution of the random effects. This is achieved by extending the ‘Nonparametric Maximum Likelihood’ towards a ‘Nonparametric profile maximum likelihood’ technique, allowing to deal with overdispersion as well as two-level data scenarios.

Download Full-text

Bayesian spatial and spatio-temporal approaches to modelling dengue fever: a systematic review

Epidemiology and Infection ◽

10.1017/s0950268818002807 ◽

2018 ◽

Vol 147 ◽

Cited By ~ 8

Author(s):

A. Aswi ◽

S. M. Cramb ◽

P. Moraga ◽

K. Mengersen

Keyword(s):

Dengue Fever ◽

Random Effects ◽

Mixed Model ◽

Linear Mixed Model ◽

Assessment Tool ◽

Temporal Dynamics ◽

Random Effect ◽

Temperature And Precipitation ◽

Generalised Linear Mixed Model ◽

Spatio Temporal

AbstractDengue fever (DF) is one of the world's most disabling mosquito-borne diseases, with a variety of approaches available to model its spatial and temporal dynamics. This paper aims to identify and compare the different spatial and spatio-temporal Bayesian modelling methods that have been applied to DF and examine influential covariates that have been reportedly associated with the risk of DF. A systematic search was performed in December 2017, using Web of Science, Scopus, ScienceDirect, PubMed, ProQuest and Medline (via Ebscohost) electronic databases. The search was restricted to refereed journal articles published in English from January 2000 to November 2017. Thirty-one articles met the inclusion criteria. Using a modified quality assessment tool, the median quality score across studies was 14/16. The most popular Bayesian statistical approach to dengue modelling was a generalised linear mixed model with spatial random effects described by a conditional autoregressive prior. A limited number of studies included spatio-temporal random effects. Temperature and precipitation were shown to often influence the risk of dengue. Developing spatio-temporal random-effect models, considering other priors, using a dataset that covers an extended time period, and investigating other covariates would help to better understand and control DF transmission.

Download Full-text

Goats Given Transdermal Flunixin Meglumine Displayed Less Pain Behavior After Castration

10.1101/2020.06.04.134049 ◽

2020 ◽

Author(s):

Amanda Lee ◽

Meggan Graves ◽

Andrea Lear ◽

Sherry Cox ◽

Marc Caldwell ◽

...

Keyword(s):

Fixed Effects ◽

Mixed Model ◽

Linear Mixed Model ◽

Body Position ◽

Random Effect ◽

Pain Behavior ◽

Bout Duration ◽

Flunixin Meglumine ◽

Meat Goats ◽

Transdermal Application

AbstractPain management should be utilized with castration to reduce physiological and behavioral changes. Transdermal application of drugs require less animal management and fewer labor risks, which can occur with oral administration or injections. The objective was to determine the effects of transdermal flunixin meglumine on meat goats’ behavior post-castration. Male goats (N = 18; mean body weight ± standard deviation: 26.4 ± 1.6 kg) were housed individually in pens and randomly assigned to 1 of 3 treatments: (1) castrated, dosed with transdermal flunixin meglumine; (2) castrated, dosed with transdermal placebo; and (3) sham castrated, dosed with transdermal flunixin meglumine. Body position, rumination, and head- pressing were observed for 1 h ± 10 minutes twice daily on days −1, 0, 1, 2, and 5 around castration. Each goat was observed once every 5-minutes (scan samples) and reported as percentage of observations. Accelerometers were used to measure standing, lying, and laterality (total time, bouts, and bout duration). A linear mixed model was conducted using GLIMMIX. Fixed effects of treatment, day relative to castration, and treatment*day relative to castration and random effect of date and goat nested within treatment were included. Treatment 1 goats (32.7 ± 2.8%) and treatment 2 goats (32.5 ± 2.8%) ruminated less than treatment 3 goats (47.4 ± 2.8%, P = 0.0012). Head pressing was greater on day of castration in treatment 2 goats (P < 0.001). Standing bout duration was greatest in treatment 2 goats on day 1 post-castration (P < 0.001). Lying bout duration was greatest in treatment 2 goats on day 1 post-castration compared to treatment 1 and treatment 3 goats(P < 0.001). Transdermal flunixin meglumine improved goats’ fluidity of movement post-castration and decreased head pressing, indicating a mitigation of pain behavior.

Download Full-text

PSVII-9 Effects of a multi-species direct-fed microbial on growth performance and carcass characteristics of grow-finish pigs

Journal of Animal Science ◽

10.1093/jas/skz122.380 ◽

2019 ◽

Vol 97 (Supplement_2) ◽

pp. 216-216

Author(s):

Mariana Boscato Menegat ◽

Joel M DeRouchey ◽

Jason C Woodworth ◽

Mike D Tokach ◽

Steve S Dritz ◽

...

Keyword(s):

Growth Performance ◽

Mixed Model ◽

Linear Mixed Model ◽

Block Design ◽

Control Diet ◽

Random Effect ◽

Carcass Characteristics ◽

Experimental Unit ◽

Feeding Duration ◽

Corn Distillers

Abstract This study was conducted to determine the effects of a multi-species direct-fed microbial (DFM) product based on lactic acid bacteria and Bacillus subtilis on growth performance and carcass characteristics of grow-finish pigs. A total of 1,188 pigs (PIC 359 × 1050; initially 25.8 kg BW) were used in a 121-d growth trial with 27 pigs/pen and 22 pens/treatment. Pigs were allotted to treatments based on initial BW in a randomized complete block design. Treatments included a control diet and the control diet with added DFM (BiOWiSH Technologies Inc., Cincinnati, OH) included at 0.055% of the diet at the expense of corn. Diets were based on corn, distillers dried grains with solubles, and soybean meal and fed in four dietary phases. Data were analyzed using a linear mixed model (PROC GLIMMIX, SAS®) with treatment as fixed effect, block as random effect, and pen as experimental unit. Overall (d 0 to 121), pigs fed the control diet had greater ADG (P < 0.05) and final BW (P < 0.001) compared to pigs fed the DFM diet (Table 1). There was no evidence for differences (P > 0.05) in ADFI or G:F between treatments. The difference in final BW resulted in heavier (P < 0.05) HCW in control pigs compared to DFM pigs, but no evidence for differences (P > 0.05) was observed in carcass yield, backfat, loin depth, and percentage lean between treatments. In conclusion, the inclusion of this multi-species DFM in growing-finishing diets reduced ADG in this commercial study. This response could be related to inclusion rate, feeding duration, or other factors not identified in this study, warranting further research to characterize the effects on pig performance.

Download Full-text

Factors associated with the weight of individual primal cuts and their inter-relationship in cattle

Translational Animal Science ◽

10.1093/tas/txz134 ◽

2019 ◽

Vol 3 (4) ◽

pp. 1593-1605 ◽

Cited By ~ 6

Author(s):

Michelle M Judge ◽

Thierry Pabiou ◽

Stephen Conroy ◽

Rory Fanning ◽

Martin Kinsella ◽

...

Keyword(s):

Decision Support ◽

Fixed Effects ◽

Mixed Model ◽

Random Effect ◽

Carcass Weight ◽

Holstein Friesian ◽

Support Tools ◽

Beef Herd ◽

Fat Score ◽

The Individual

Abstract Input parameters for decision support tools are comprised of, amongst others, knowledge of the associated factors and the extent of those associations with the animal-level feature of interest. The objective of the present study was to quantify the association between animal-level factors with primal cut yields in cattle and to understand the extent of the variability in primal cut yields independent carcass weight. The data used consisted of the weight of 14 primal carcass cuts (as well as carcass weight, conformation, and fat score) on up to 54,250 young cattle slaughtered between the years 2013 and 2017. Linear mixed models, with contemporary group of herd-sex-season of slaughter as a random effect, were used to quantify the associations between a range of model fixed effects with each primal cut separately. Fixed effects in the model were dam parity, heterosis coefficient, recombination loss, a covariate per breed representing the proportion of Angus, Belgian Blue, Charolais, Jersey, Hereford, Limousin, Simmental, and Holstein–Friesian and a three-way interaction between whether the animal was born in a dairy or beef herd, sex, and age at slaughter, with or without carcass weight as a covariate in the mixed model. The raw correlations among all cuts were all positive varying from 0.33 (between the bavette and the striploin) to 0.93 (between the topside and knuckle). The partial correlation among cuts, following adjustment for differences in carcass weight, varied from −0.36 to 0.74. Age at slaughter, sex, dam parity, and breed were all associated (P < 0.05) with the primal cut weight. Knowledge of the relationship between the individual primal cuts, and the solutions from the models developed in the study, could prove useful inputs for decision support systems to increase performance.

Download Full-text

Misspecification of the random effect structure: Implications for the linear mixed model

10.17077/67hl-x2l9 ◽

2020 ◽

Author(s):

Brandon LeBeau

Keyword(s):

Fixed Effects ◽

Serial Correlation ◽

Mixed Model ◽

Linear Mixed Model ◽

Type I Error ◽

Random Effect ◽

Correlated Data ◽

Unbiased Estimation ◽

Type I ◽

Nested Data

<p>The linear mixed model is a commonly used model for longitudinal or nested data due to its ability to account for the dependency of nested data. Researchers typically rely on the random effects to adequately account for the dependency due to correlated data, however serial correlation can also be used. If the random effect structure is misspecified (perhaps due to convergence problems), can the addition of serial correlation overcome this misspecification and allow for unbiased estimation and accurate inferences? This study explored this question with a simulation. Simulation results show that the fixed effects are unbiased, however inflation of the empirical type I error rate occurs when a random effect is missing from the model. Implications for applied researchers are discussed.</p>

Download Full-text

Exercise-induced hypoalgesia in women with varying levels of menstrual pain

Scandinavian Journal of Pain ◽

10.1515/sjpain-2018-0020 ◽

2018 ◽

Vol 18 (2) ◽

pp. 303-310 ◽

Cited By ~ 4

Author(s):

Mervyn Travers ◽

Penny Moss ◽

William Gibson ◽

Dana Hince ◽

Sheree Yorke ◽

...

Keyword(s):

Fixed Effects ◽

Mixed Model ◽

Linear Mixed Model ◽

Random Effect ◽

Activity Levels ◽

Menstrual Pain ◽

Isometric Handgrip ◽

Pressure Pain Thresholds ◽

Significant Difference ◽

Exercise Induced

Abstract Background and aims: Exercise-induced hypoalgesia (EIH) is a well-established phenomenon in pain-free individuals that describes a decrease in pain sensitivity after an acute bout of exercise. The EIH response has been demonstrated to be sub-optimal in the presence of persisting pain. Menstrual pain is a common recurrent painful problem with many women experiencing high levels of pain each cycle. However, the EIH response has not been examined in a cohort of women with high levels of menstrual pain. This research aimed to examine whether EIH manifests differently in women with varying levels of menstrual pain. The primary hypothesis was that women with high levels of menstrual pain would demonstrate compromised EIH. Secondary aims were to explore relationships between EIH and emotional state, sleep quality, body mass index (BMI) or physical activity levels. Methods: Pressure pain thresholds (PPT) were measured in 64 participants using a digital handheld algometer before and after a submaximal isometric-handgrip exercise. EIH index was compared between low (VAS 0–3), moderate (VAS 4–7) and high (VAS 8–10) pain groups, using a linear mixed model analysis with participant as a random effect, and site, menstrual pain category and the interaction between the two, as fixed effects. Results: EIH was consistently induced in all groups. However, there was no statistically significant difference between the pain groups for EIH index (p=0.835) or for any co-variates (p>0.05). Conclusions: EIH was not found to differ between women who report regular low, moderate or high levels of menstrual pain, when measured at a point in their menstrual cycle when they are pain free. Implications: This study provides insight that EIH does not vary in women with differing levels of menstrual pain when they are not currently experiencing pain. The current findings indicate that, although menstrual pain can involve regular episodes of high pain levels, it may not be associated with the same central nervous system dysfunctions as seen in sustained chronic pain conditions.

Download Full-text

Hand Hygiene With Alcohol-Based Hand Rub: How Long Is Long Enough?

Infection Control and Hospital Epidemiology ◽

10.1017/ice.2017.25 ◽

2017 ◽

Vol 38 (5) ◽

pp. 547-552 ◽

Cited By ~ 25

Author(s):

Daniela Pires ◽

Hervé Soule ◽

Fernando Bellissimo-Rodrigues ◽

Angèle Gayet-Ageron ◽

Didier Pittet

Keyword(s):

Hand Hygiene ◽

Mixed Model ◽

Linear Mixed Model ◽

Random Effect ◽

Healthcare Personnel ◽

Infection Prevention And Control ◽

Bacterial Reduction ◽

Experimental Conditions ◽

Bacterial Counts ◽

Hand Rubbing

BACKGROUNDHand hygiene is the core element of infection prevention and control. The optimal hand-hygiene gesture, however, remains poorly defined.OBJECTIVEWe aimed to evaluate the influence of hand-rubbing duration on the reduction of bacterial counts on the hands of healthcare personnel (HCP).METHODSWe performed an experimental study based on the European Norm 1500. Hand rubbing was performed for 10, 15, 20, 30, 45, or 60 seconds, according to the WHO technique using 3 mL alcohol-based hand rub. Hand contamination with E. coli ATCC 10536 was followed by hand rubbing and sampling. A generalized linear mixed model with a random effect on the subject adjusted for hand size and gender was used to analyze the reduction in bacterial counts after each hand-rubbing action. In addition, hand-rubbing durations of 15 and 30 seconds were compared to assert non-inferiority (0.6 log10).RESULTSIn total, 32 HCP performed 123 trials. All durations of hand rubbing led to significant reductions in bacterial counts (P<.001). Reductions achieved after 10, 15, or 20 seconds of hand rubbing were not significantly different from those obtained after 30 seconds. The mean bacterial reduction after 15 seconds of hand rubbing was 0.11 log10 lower (95% CI, −0.46 to 0.24) than after 30 seconds, demonstrating non-inferiority.CONCLUSIONSHand rubbing for 15 seconds was not inferior to 30 seconds in reducing bacterial counts on hands under the described experimental conditions. There was no gain in reducing bacterial counts from hand rubbing longer than 30 seconds. Further studies are needed to assess the clinical significance of our findings.Infect Control Hosp Epidemiol 2017;38:547–552

Download Full-text

Linear mixed model for heritability estimation that explicitly addresses environmental variation

Proceedings of the National Academy of Sciences ◽

10.1073/pnas.1510497113 ◽

2016 ◽

Vol 113 (27) ◽

pp. 7377-7382 ◽

Cited By ~ 42

Author(s):

David Heckerman ◽

Deepti Gurdasani ◽

Carl Kadie ◽

Cristina Pomilla ◽

Tommy Carstensen ◽

...

Keyword(s):

Mixed Model ◽

Linear Mixed Model ◽

Random Effect ◽

Simulated Data ◽

Spatial Location ◽

Liver Function Tests ◽

Narrow Sense Heritability ◽

Heritability Estimation ◽

Genome Wide Data ◽

Estimate Heritability

The linear mixed model (LMM) is now routinely used to estimate heritability. Unfortunately, as we demonstrate, LMM estimates of heritability can be inflated when using a standard model. To help reduce this inflation, we used a more general LMM with two random effects—one based on genomic variants and one based on easily measured spatial location as a proxy for environmental effects. We investigated this approach with simulated data and with data from a Uganda cohort of 4,778 individuals for 34 phenotypes including anthropometric indices, blood factors, glycemic control, blood pressure, lipid tests, and liver function tests. For the genomic random effect, we used identity-by-descent estimates from accurately phased genome-wide data. For the environmental random effect, we constructed a covariance matrix based on a Gaussian radial basis function. Across the simulated and Ugandan data, narrow-sense heritability estimates were lower using the more general model. Thus, our approach addresses, in part, the issue of “missing heritability” in the sense that much of the heritability previously thought to be missing was fictional. Software is available at https://github.com/MicrosoftGenomics/FaST-LMM.

Download Full-text