Simulation of Subgroup Analysis Methods with Longitudinal Data Containing Measurement Errors and Missingness / 中国卫生统计

Objective To develop a subgroup analysis method that can simultaneously deal with longitudinal data containing measurement errors and dropouts.Methods Subgroup analysis was carried out based on a threshold regression model.A new generalized unbiased estimation equation is constructed by using the independence between repeated measurements to deal with measurement errors and introducing an inverse probability weighting matrix to deal with missing response.Results The computer stochastic simulation shows that the proposed estimation method is effective in dealing with measurement errors and dropouts,and has smaller bias and mean square error than the generalized estimation equation method without correcting measurement errors or dropouts.Conclusion In subgroup analysis,when there are measurement errors in covariables and missing values in response variables,it is usually necessary to deal with the measurement errors and missing values in order to obtain reliable parameter estimation.

Application of Linear Mixed Model Tree in Longitudinal Trajectory of Body Mass Index / 中国卫生统计

Objective To understand the trajectory and classification of adult body mass index(BMI)in Jiangsu Province.Methods Based on China Health and Nutrition Survey,this study used the linear mixed model tree to explore the trajectory and classification of BMI of people aged 18-65 in Jiangsu Province.Results The linear mixed model tree had 13 nodes and the depth was 6.The classification nodes were baseline BMI,average calorie intake and baseline age.Conclusion The linear mixed model tree can identify the trajectory of BMI and expand the research method of longitudinal data.

Proposal for a new non-linear model to describe growth curves

This study was developed with longitudinal data measurements of Norfolk rabbits from birth to 119 days of age to estimate the average growth curve, with the primary objective of proposing a non-linear model. It also selected the most appropriate sigmoidal model to describe the growth of Norfolk rabbits. The adjustments provided by the logistic, von Bertalanffy, Gompertz, Brody, Richards, and proposed models were compared. The parameters were estimated using the "nls" function of the "stats" package in R software, the least-squares method, and the Gauss-Newton convergence algorithm. The goodness-of-fit comparison was based on the following criteria: adjusted coefficient of determination (), mean square error (MSE), mean absolute deviation (MAD), Akaike information criterion (AIC), and Bayesian information criterion (BIC). Cluster analysis helped select and classify the non-linear growth models, considering the other goodness-of-fit criteria results. The proposed non-linear, von Bertalanffy, Gompertz, and Richards models described the growth curve of Norfolk rabbits satisfactorily, providing parameters with practical interpretations. The goodness-of-fit criteria showed that the proposed and von Bertalanffy models best represented the growth of rabbits.

Relationship between reduction of urban PM2.5 and mental health of middle-aged and elderly people after China's Air Pollution Prevention and Control Action Plan / 环境与职业医学

Background Epidemiological evidence indicates an association of particulate matter with depression and cognitive performance. From 2013 to 2017, China implemented the Air Pollution Prevention and Control Action Plan to reduce particulate matter concentration. There are few studies on the relationship between the decrease of particulate matter concentration and the improvement of mental health in middle-aged and elderly people. Objective To analyze the relationship between the decrease of city-level particulate matter concentration and the improvement of depression and cognitive function in the middle-aged and elderly population after the implementation of the Air Pollution Prevention and Control Action Plan. Methods Using the China Health and Retirement Longitudinal Study (CHARLS) data in 2011 and 2018, this study applied longitudinal data clustering technology to group cities based on the actual response of each city to the policy (the dynamic change trajectory of PM2.5 in each city during the study period); the higher the degree of response, the greater the reduction of PM2.5 concentration in the city. We assigned participants to three groups with different degrees of response to the policy, including low-response group A as the control group, medium-response intervention group B, and high-response intervention group C. A difference-in-differences (DID) model was used to estimate the influence of PM2.5 decline on the depression and cognitive function among middle-aged and elderly people in China by considering potential individual and city-level time-varying confounders. Depression was measured using the 10-item Center for the Epidemiological Studies of Depression Scale (CES-D10) (10 questions, total score range 0-30) with higher score representing higher severity of depression. Cognitive function was evaluated with reference to the international cognitive function test questionnaire for the middle-aged and the elderly which was further categorized into two dimensions of memory and cognitive status and included 31 questions with a score range of 0-31; the higher the score, the better the cognitive function. Samples with relatively complete outcomes were selected for analysis, including 10729 people in depression analysis and 4510 people in cognitive analysis. Results The longitudinal clustering results indicated that the urban groups with the highest decline of PM2.5 concentrations (high-response group C) had the highest baseline PM2.5 concentrations, mainly in Beijing-Tianjin-Hebei region, Chengdu-Chongqing metropolitan area, and Wuhan metropolitan area. In 2011, no significant differences were observed in depression and cognitive function among the three groups of middle-aged and elderly populations (Kruskal Wallis test: Pdepression=0.864, Pcognition=0.239). Significant differences were found in depression and cognitive function in both low-response group A and medium-response group B in 2018 compared to 2011 (paired Wilcox test, all P<0.001). However, in the high-response group C, there was no significant difference in depression in 2018 compared to 2011 (P=0.195), while a significant difference was detected in cognitive function (P=0.006). As PM2.5 concentrations decreased, the DID model showed that the depression of the middle-aged and elderly people in the high-response group C decreased by 7.55% (95%CI: 2.83%-12.03%), and the cognitive function improved by 2.70% (95%CI: 0.25%-5.22%) compared with the low-response group A. However, no intervention effect was observed in group B with moderate response level compared with group A with low response level. Conclusion After the implementation of the Air Pollution Prevention and Control Action Plan policy, the decrease of PM2.5 concentration has an ameliorative effect on the depression and cognitive function of middle-aged and elderly people in China. Given the aging population and the increasing burden of mental-related diseases in China, the promotion of environmental air pollution control has important public health implications.

Application of the Joint Latent Class Model in the Reversion of Mild Cognitive Impairment to Normal Cognition / 中国卫生统计

Objective The joint latent class model(JLCM)was applied to study the reversion of mild cognitive impairment(MCI)to normal cognition(NC),providing a methodology reference for the longitudinal study of chronic diseases.Methods The JLCM model consisted of three sub-models:the latent class sub-model adopted multi-class logistic regression to estimate the latent classes;the longitudinal process sub-model adopted linear mixed model to describe the longitudinal cognitive measurement;and the survival process sub-model adopted proportional risk model to fit the cognitive process.Goodness of fit was based on information criterion,posterior classification,comparison between predicted value and observed value and conditional independence assumption.Results The MCI population was divided into two latent classes based on the JLCM model,including 783(92.88%)and 60(7.12%),respectively.The latent class sub-model showed that gender,marital status,APOE4 and FAQ had statistical significance for the determination of latent classes.The longitudinal process sub-model showed male(β=-0.685,95%CI:-1.144,-0.226),being single(β=0.743,95%CI:0.200,1.286),not carrying APOE4(β=-1.201,95%CI:-1.636,-0.766)and without functional impairment(β=-1.868,95%CI:-2.095,-1.641)had higher MMSE scores.The survival process sub-model showed that high education level(HR=1.264,95%CI:1.134,1.395),being single(HR=1.593,95%CI:1.286,1.899)and not carrying APOE4(HR=0.453,95%CI:0.043,0.862)were protective factors for reversion to NC in MCI patients.Conclusion JLCM model has good application value in the study of chronic diseases such as cognitive impairment.

A SIMPLE GIBBS SAMPLER FOR THE STATE ESTIMATION IN WIRELESS COMMUNICATIONS

In the recent years, wireless communications are extremely useful in many disciplines including health monitoring, environment monitoring, signal processing etc. State estimation and prediction are quite challenging tasks in wireless communications. Traditionally, in the literature, dynamic state-space models have been used for the state estimation and predic- tion purpose. The estimation method is based on Kalman-Filter which is computationally demanding. In this work, we consider computationally simpler Gibbs sampler algorithm for the state estimation. We consider three different cases, (i) continuous state values, (ii) binary (0/1) state values, and (iii) categorical state values with more than two categories. We consider a simple linear model for the prediction purpose, and the underlying regression coefficients are estimated by Gibbs sampler. We compute the misclassification proportions for assessing the practical usefulness of our estimation approach. Areal dataset where 200 wireless sensor nodes are used for measuring the temperature of a chamber is analysed in this work

Multiple imputation of missing data in clinical longitudinal studies and its sensitivity analyses / 中国临床药理学与治疗学

AIM: To guide the multiple imputation of missing data in clinical longitudinal studies and its sensitivity analyses, and highlight the importance of sensitivity analyses by taking the clinical trial of Qizhitongluo Capsule in treating ischemic stroke as an example. METHODS: To implement PROC MI process in SAS to perform multiple imputation and its sensitivity analysis. RESULTS: In the example, after multiple imputation, improvements in lower limb motor scores of the Qizhitongluo group were greater than those of the placebo group (all P<0.01), and the results of two sensitivity analyses under "missing not at random" were consistent with those under "missing at random". CONCLUSION: Multiple imputations combined with sensitivity analyses can ensure a robust result. It is recommended that clinical researchers perform sensitivity analyses after filling missing data.

Trajetórias educacionais como evidência da qualidade da educação básica brasileira / Educational trajectories as evidence of the quality of Brazilian basic education / Trayectorias educativas como evidencia de la calidad de la educación básica brasileña

Segundo a Constituição Federal brasileira, o direito à educação significa a garantia do acesso à escola, onde a população deve permanecer durante o tempo legalmente determinado, a fim de adquirir os conhecimentos e habilidades necessários para a vida adulta plena. Este artigo tem como foco a análise da permanência dos estudantes na escola, condição essencial para a aprendizagem e representada por suas trajetórias educacionais. Dados longitudinais do Censo Escolar, de 2007 a 2015, foram utilizados para analisar as trajetórias educacionais classificadas em três tipos: regular; com poucas irregularidades; e com muitas irregularidades. Os resultados mostram que alguns grupos sociais são muito mais propensos a ter uma trajetória regular e outros a ter repetência ou a ficar fora da escola em alguns anos. Uma proporção alta de meninos de cor/raça preta tem trajetórias com muitas irregularidades. O artigo mostra que o Índice de Desenvolvimento da Educação Básica (Ideb) - o indicador oficial para medir a qualidade da educação nos municípios - pode atingir um valor alto, mesmo que muitos estudantes tenham trajetórias irregulares. Isso evidenciou o efeito da seletividade embutido nesse indicador. Em conclusão, este trabalho sugere a incorporação das trajetórias educacionais no monitoramento da qualidade da educação nos municípios.

According to the Federal Brazilian Constitution, the right to education means that all children had access to a school, and remained there for the legally determined time and, acquired the knowledge and skills necessary for full adult life. This paper focuses on the analysis of students' permanence in school, an essential condition for learning represented here by their educational trajectories. The School Census longitudinal data, from 2007 to 2015 were used to analyze the student's educational trajectories classified into three types: regular; with few irregularities; and with many irregularities. The results show that some social groups are much more prone to having a regular trajectory, and others to having grade repetition or being out of school after few years. A large proportion of black boys have trajectories with many irregularities. In addition, this paper shows that the Basic Education Development Index (IDEB, acronym in Portuguese) - the official index for measuring the quality of education in the municipalities - can reach a high value even if many students have irregular trajectories. This evidenced the selectivity effect embedded in this indicator. This paper suggests the incorporation of educational trajectories into the indicator to monitor the quality of education in Brazilian municipalities.

De acuerdo con la Constitución Federal brasileña, el derecho a la educación significa la garantía de acceso a la escuela, donde los estudiantes deben permanecer por el tiempo legalmente determinado para adquirir los conocimientos y habilidades necesarios para la vida adulta plena. Este artículo se centra en el análisis de la permanencia de los estudiantes en la escuela, condición esencial para el aprendizaje, y representada por sus trayectorias educativas. Datos longitudinales de los censos escolares desde 2007 hasta 2015 se utilizaron para estudiar las trayectorias educativas del alumno, que se clasificaron en tres tipos: regular, con pocas irregularidades, con muchas irregularidades. Los resultados muestran que algunos grupos sociales son mucho más propensos a tener una trayectoria regular y otros a repetir el grado o quedar fuera de la escuela por algunos años. Una gran proporción de estudiantes negros tiene trayectorias con muchas irregularidades. Además, este trabajo muestra que el Índice de Desarrollo de la Educación Básica (IDEB) —índice oficial para medir la calidad de la educación en los municipios— puede alcanzar un valor alto a pesar de que muchos estudiantes tienen trayectorias irregulares. Esto evidenció el efecto de selectividad incorporado en este indicador. Este trabajo sugiere la incorporación de las trayectorias educativas en el indicador de seguimiento de la calidad de la educación en los municipios brasileños.

Association between total cholesterol and type 2 diabetes: a prospective cohort study / 中华疾病控制杂志

@#Objective To explore the association between total cholesterol and type 2 diabetes ( T2DM) ． Methods Non－diabetic people who aged 20 to 90 years at the baseline and who had physical examination more than 2 times were screened． Comparisons of the baseline characteristics were conducted with Student－t test or Pearson chi－square test． Generalized estimating equation ( GEE) was used to analyze the effect of total cholesterol of quantiles groups ( 2．10－ mmol /L，4．16－ mmol /L，4．76－ mmol /L and 5．42 －13．29 mmol /L) to type 2 diabetes． Results The cohort with an average age of 3．53 years per person in- cluded 12 928 subjects and 45 626 person－years． During the follow－up，447 cases of new－onset diabetes occurred and the incidence density was 9. 80‰． The high incidence of type 2 diabetes increased with the increase of total cholesterol． After adjusting the factors including age，high density lipoprotein，hypertension and obesity，based on the 2. 10－ mmol /L group，the relative risk ( RR) of the 4. 16－ mmol /L，4. 76－ mmol/L and 5. 42－13. 29 mmol /L group were 1. 24( 95% CI: 0. 83－1. 86) ，1. 75 ( 95% CI: 1. 19－2. 56) and 3. 60( 95% CI: 2. 51－5. 17) ，respectively． Conclusions Total cholesterol is associated with type 2 diabetes，and as the total cholesterol increases，the risk of developing type 2 diabetes increases．

Analyzing weight evolution in mice infected by Trypanosoma cruzi

Concerning the specificities of a longitudinal study, the trajectories of a subject's mean responses not always present a linear behavior, which calls for tools that take into account the non-linearity of individual trajectories and that describe them towards associating possible random effects with each individual. Generalized additive mixed models (GAMMs) have come to solve this problem, since, in this class of models, it is possible to assign specific random effects to individuals, in addition to rewriting the linear term by summing unknown smooth functions, not parametrically specified, then using the P-splines smoothing technique. Thus, this article aims to introduce this methodology applied to a dataset referring to an experiment involving 57 Swiss mice infected by Trypanosoma cruzi, which had their weights monitored for 12 weeks. The analyses showed significant differences in the weight trajectory of the individuals by treatment group; besides, the assumptions required to validate the model were met. Therefore, it is possible to conclude that this methodology is satisfactory in modeling data of longitudinal sort, because, with this approach, in addition to the possibility of including fixed and random effects, these models allow adding complex correlation structures to residuals.

Introduction of joint model and its applications in medical research / 中华流行病学杂志

In medical follow-up studies, longitudinal data and survival data are often accompanied and associated with each other, thus respective analysis of longitudinal and survival data might lead to biased results. Joint model can correct deviations, improve the efficiency of parameter estimation and provide effective inferences by simultaneously processing longitudinal and survival data. It is a popular method in medical research. Joint model has made much progress, whereas the literature about the joint model and its application is limited in China. This paper summarizes the main idea, basic framework, parameter estimation methods of random effect joint model and introduces the analysis on AIDS data set based on the R software package 'JM’ to clarify the advantages of the joint model in processing medical follow-up data and promote the use of the joint model in clinical research.

Changes in quality of life and acceptance of disability of burn patients in rehabilitation treatment stage and the influencing factors / 中华烧伤杂志

Objective@#To explore the development trajectories of quality of life and acceptance of disability of burn patients in the rehabilitation treatment stage and the influencing factors.@*Methods@#Totally 207 burn patients, including 157 males and 50 females, aged (40±13) years, who were in the rehabilitation treatment stage were selected by convenient sampling method from October 2016 to July 2017 in the Department of Burns of Fujian Medical University Union Hospital for this longitudinal study. At discharge and 1, 3, and 6 months after discharge, the patient′s quality of life and acceptance of disability were scored using the Burn Specific Health Scale-Brief and Chinese Version of Acceptance of Disability Scale-Revised respectively. Taking the intercept, the slope, and the curve slope as latent variables, the latent second growth curve model was constructed for the quality of life and the acceptance of disability. The robust maximum likelihood estimation (MLR) method was used to estimate the mean, the variance, and the covariance, so as to analyze the discharge level, the growth rate, the acceleration, and the correlation among them. Taking the acceptance of disability, the gender, the cause of burn, the severity of burn, the existence of complications, the payment way, and the education level as covariates, the latent second growth curve model was constructed for the quality of life. The MLR method was used to estimate the influence of covariates on the discharge level, the growth rate, and the acceleration of the quality of life.@*Results@#At discharge and 1, 3, and 6 months after discharge, the quality of life scores of patients were (102±36), (111±36), (118±37), and (122±37) points respectively, and the acceptance of disability scores were (73±17), (75±17), (77±17), and (78±18) points respectively. The estimated mean intercept of the quality of life and the acceptance of disability were 101.680 and 72.993 respectively at discharge, both of which showed a curve increasing trend in 1, 3, and 6 months after discharge (estimated mean slope=11.024, 3.086, t=15.376, 7.476, P<0.01), and the increasing rate (acceleration) gradually slowed down (estimated mean curve slope=-1.393, -0.426, t=-13.339, -4.776, P<0.01). There were significant individual differences in the discharge level and the acceleration of quality of life of patients (estimated intercept variance=1 174.527, t=9.332; estimated curve slope variance=2.379, t=6.402; P<0.01). There were significant individual differences in the discharge level, the growth rate, and the acceleration of patients′ acceptance of disability (estimated intercept variance=267.017, t=9.262; estimated slope variance=32.264, t=2.356; estimated curve slope variance=0.882, t=2.939; P<0.05 or P<0.01). There was no significant correlation among the discharge level, the growth rate, and the acceleration of the quality of life and those of the acceptance of disability of patients (estimated intercept and slope=37.273, -1.457, t=0.859, -0.131; estimated intercept and curve slope=-6.712, -0.573, t=-1.089, -0.248; estimated slope and curve slope=-5.494, -5.988, t=-0.930, -2.512; P>0.05). Among the time-constant covariates, only the severity of burn and the presence of complications had a significant impact on the quality of life of patients at discharge (estimated intercept=-10.721, 5.522, t=-6.229, 1.977, P<0.05 or P<0.01). At discharge and 1, 3, and 6 months after discharge, the level of acceptance of disability had a positive impact on the quality of life of patients (standardized regression coefficient=0.616, 0.669, 0.681, 0.678, t=18.874, 21.660, 22.824, 22.123, P<0.01).@*Conclusions@#The initial levels of quality of life and acceptance of disability of burn patients in the rehabilitation treatment stage are relatively low, both with a curve increasing trend over time, and the increasing rate gradually slows down. Patients with complications and serious burns have poor quality of life at discharge, while the acceptance of disability has a positive impact on the quality of life.

Effect of Long-Term Systolic Blood Pressure Trajectory on Kidney Damage in the Diabetic Population: A Prospective Study in a Community-Based Chinese Cohort / 中华医学杂志(英文版)

<p><b>Background</b>Previous studies have shown that hypertension is an important factor contributing to the occurrence and progression of diabetic kidney damage. However, the relationship between the patterns of blood pressure (BP) trajectory and kidney damage in the diabetic population remains unclear. This prospective study investigated the effect of long-term systolic BP (SBP) trajectory on kidney damage in the diabetic population based on an 8-year follow-up community-based cohort.</p><p><b>Methods</b>This study included 4556 diabetic participants among 101,510 participants. BP, estimated glomerular filtration rate (eGFR), and urinary protein were measured every 2 years from 2006 to 2014. SBP trajectory was identified by the censored normal modeling. Five discrete SBP trajectories were identified according to SBP range and the changing pattern over time. Kidney damage was evaluated through eGFR and urinary protein value. A multivariate logistic regression model was used to analyze the influence of different SBP trajectory groups on kidney damage.</p><p><b>Results</b>We identified five discrete SBP trajectories: low-stable group (n = 864), moderate-stable group (n = 1980), moderate increasing group (n = 609), elevated decreasing group, (n = 679), and elevated stable group (n = 424). The detection rate of kidney damage in the low-stable group (SBP: 118-124 mmHg) was the lowest among the five groups. The detection rate of each kidney damage index was higher in the elevated stable group (SBP: 159-172 mmHg) compared with the low-stable group. For details, the gap was 4.14 (11.6% vs. 2.8%) in eGFR <60 ml·min·1.73 m and 3.66 (17.2% vs. 4.7%), 3.38 (25.0% vs. 7.4%), and 1.8 (10.6% vs. 5.9%) times in positive urinary protein, eGFR <60 ml·min·1.73 m and/or positive urinary protein, and eGFR decline ≥30%, respectively (P < 0.01).</p><p><b>Conclusion</b>An elevated stable SBP trajectory is an independent risk factor for kidney damage in the diabetic population.</p>

The latent variable growth curve model of longitudinal data and its implementation in Mplus / 中华流行病学杂志

To discuss the latent variable growth curve model of longitudinal data and give its implementation method in Mplus.The application of Mplus software has been used to deal with the longitudinal data of mental health status of college students in an university.Results show that the model can process the longitudinal data with latent variables,which can compare the differences of the overall development trend and individual development,also taking a covariate into the model to improve the effect of model fitting.Using Mplus software to process the longitudinal data with latent variables,the program is simple and easy to operate.This study provides the latent variable growth curve model of longitudinal data and its procedure of implementation in Mplus,and the statistical methodology guidance and reference for practical applications of epidemiological cohort study.

The latent variable growth curve model of longitudinal data and its implementation in Mplus / 中华流行病学杂志

To discuss the latent variable growth curve model of longitudinal data and give its implementation method in Mplus.The application of Mplus software has been used to deal with the longitudinal data of mental health status of college students in an university.Results show that the model can process the longitudinal data with latent variables,which can compare the differences of the overall development trend and individual development,also taking a covariate into the model to improve the effect of model fitting.Using Mplus software to process the longitudinal data with latent variables,the program is simple and easy to operate.This study provides the latent variable growth curve model of longitudinal data and its procedure of implementation in Mplus,and the statistical methodology guidance and reference for practical applications of epidemiological cohort study.

Joint Modeling of Multivariate Longitudinal Data and Its Application / 世界科学技术-中医药现代化

Multiple outcomes measured repeatedly for the same subject are common in longitudinal observation.If we use the approach by analyzing each outcome separately,it may lead to wrong conclusions due to the failure of accounting for joint evolution of different outcomes.To adequately capture the interdependence among multiple outcomes,we proposed a joint modeling for multivariate longitudinal data by constructing a linear mixed-effects model for each outcome and accommodating the relationship among multiple outcomes through correlation in random effects.Maximum likelihood method was adopted to estimate parameters in this model.The application of this method was demonstrated through the analysis of stroke data.

Ronda clínica y epidemiológica. Análisis de datos longitudinales / Clinical and epidemiological round. Longitudinal data analysis / Ronda clínica e epidemiológica. Análise de dados longitudinais

El denominado análisis de datos longitudinales (ADL) se refiere a los métodos para evaluar de manera apropiada las medidas de un mismo sujeto que se repiten en el tiempo. El ADL es una herramienta adecuada para entender indicadores de cambio en procesos de salud y enfermedad y para la evaluación del efecto de diversas intervenciones terapéuticas. Se presentan los principales modelos de ADL, sus ventajas y algunos ejemplos recientes de la literatura médica.

Longitudinal data analysis (LDA) refers to the methods designed to evaluate repeated measurements within an individual. LDA is an appropriate tool to address the process of change in health and disease and also to evaluate the efficacy of interventions. We present the main LDA models as well as their advantages and some clinical examples from recent medical literature.

A denominada análise de dados longitudinais (ADL) refere-se aos métodos para avaliar de maneira apropriada as medidas de um mesmo sujeito que se repetem no tempo. O ADL é uma ferramenta adequada para entender indicadores de mudança em processos de saúde e doença e para a avaliação do efeito de diversas intervenções terapêuticas. Apresentam-se os principais modelos de ADL, suas vantagens e alguns exemplos recentes da literatura médica.

Seleção da matriz de variância-covariância residual na análise de ensaios varietais com medidas repetidas em cana-de-açúcar / Selection of the residual variance-covariance matrix in the analysis of varietal trials with repeated measures in sugarcane

Este trabalho objetivou avaliar diferentes estruturas da matriz de variâncias e covariâncias residual (Σ), quanto ao ajustamento de dados longitudinais via modelos mistos, em experimentos varietais de cana-de-açúcar. A seleção adequada desta matriz garante a escolha de um modelo mais representativo dos dados. Em cada modelagem, variou-se ainda a suposição associada aos efeitos de tratamentos (variedades), como fixos e aleatórios. Quatro ensaios varietais, conduzidos entre 2005 e 2009, em três localidades do Estado de Goiás, foram considerados. Cada experimento foi delineado em blocos casualizados com três ou quatro repetições. A variável resposta analisada foi toneladas de colmos por hectare (TCH). Para avaliar a qualidade de ajustamento dos modelos, foram utilizados o critério de informação de Akaike (AIC) e o teste da razão de verossimilhanças. Este último foi utilizado apenas para comparar modelos hierárquicos, tomados dois a dois. Constatou-se que as análises pelo modelo univariado clássico de parcelas subdivididas oscilaram entre as piores ou entre aquelas de ajustes apenas medianos. As estruturas da matriz Σ com os melhores ajustamentos variaram entre os ensaios, com destaque para a matriz não-estruturada. Tais resultados revelam que a estrutura de erros independentes, em geral, não se mostra adequada para esse tipo de análise e, também, que não é seguro definir previamente uma estrutura específica de Σ para tais ensaios. Pequenas alterações foram observadas na classificação das estruturas ao se assumirem os efeitos de tratamentos como fixos ou aleatórios; porém, sem efeito importante na classificação das melhores estruturas em cada ensaio.

This study aimed to evaluate different residual structures of variance-covariance matrix (Σ), regarding the fitting of longitudinal data via mixed models in variety trials of sugarcane. The adequate choice of this matrix provides most representative models to the data. In each model was also evaluated the effects of treatments (varieties), either as fixed or as random. Four trials were carried out in three locations in the Goiás State, Brazil, from 2005 to 2009. Each experiment was designed in randomized complete block with three or four repetitions. The response variable analyzed was tons of stalks per hectare (TCH). The goodness of fitting of the different models to the data was assessed by Akaike information criterion (AIC) and by likelihood ratio test (LRT). This last statistic was used only to compare nested models, two by two. It was observed that classic model in split-plot design ranged among the worst or with just median adjustments. The structures of Σ matrix with the best fittings to the data varied among trials, with outstanding for the unstructured matrix. These results show that the structure of independent errors, in general, is not adequate for these analyses, and a prior definition of the co-variance structure can lead to unreliable results for these trials. Small changes were observed in the ranking of these structures by assuming the treatment effects as fixed or random, however, without significant effects on the ranking of the best structures in each trial.

Competing risk model based study of outcomes of mild cognitive impairment of seniors / 中华流行病学杂志

Objective To introduce the competing risk model into outcome prediction of mild cognitive impairment (MCI) of seniors and to explore influencing factors for the prognosis of MCI to Alzheimer' s disease (AD).Methods Data were collected from six follow-up visits to 600 seniors from communities in Taiyuan city,which were conducted at an interval of six months from October 2010 to May 2013.MCI state was defined as the transient state,AD and death before AD as two absorbing states (death before AD in which was regarded as a competing risk event),building the competing risk model to identify the model parameters,and to explore influencing factors on MCI prognosis to AD.In the meantime,the 3-year MCI-AD transition probability was estimated based on the multi-state Markov model.Results Based on screening with the multivariate competing risk model analysis,factors such as higher age (estimate HR=1.56,95％CI:1.01-2.39),female gender (HR=1.72,95％CI:1.02-2.92),higher education(HR=0.64,95％CI:0.41-1.00),reading frequently (HR=0.57,95％CI:0.32-0.99),hypertension (HR=3.43,95％CI:1.08-10.85) and high SBP(HR=1.67,95％CI:1.04-2.66),were statistically significant for transition from MCI to AD in three years.3-year MCI-AD transition probability was 10.7％(95％CI:8.6％-13.2％).Conclusion Age,gender,education,reading and blood pressure were the influencing factors for the prognosis of MCI to AD.Competing risk model was advantageous in studying longitudinal data with multiple potential outcomes.

A school-level longitudinal study of clinical performance examination scores / 한국의학교육

PURPOSE: This school-level longitudinal study examined 7 years of clinical performance data to determine differences (effects) in students and annual changes within a school and between schools; examine how much their predictors (characteristics) influenced the variation in student performance; and calculate estimates of the schools' initial status and growth. METHODS: A school-level longitudinal model was tested: level 1 (between students), level 2 (annual change within a school), and level 3 (between schools). The study sample comprised students who belonged to the CPX Consortium (n=5,283 for 2005~2008 and n=4,337 for 2009~2011). RESULTS: Despite a difference between evaluation domains, the performance outcomes were related to individual large-effect differences and small-effect school-level differences. Physical examination, clinical courtesy, and patient education were strongly influenced by the school effect, whereas patient-physician interaction was not affected much. CONCLUSION: Student scores are influenced by the school effect (differences), and the predictors explain the variation in differences, depending on the evaluation domain.