1.Analysis of binary classification repeated measurement data with GEE and GLMMs using SPSS software.
Shengli AN ; Yanhong ZHANG ; Zheng CHEN
Journal of Southern Medical University 2012;32(12):1777-1780
OBJECTIVETo analyze binary classification repeated measurement data with generalized estimating equations (GEE) and generalized linear mixed models (GLMMs) using SPSS19.0.
METHODSGEE and GLMMs models were tested using binary classification repeated measurement data sample using SPSS19.0.
RESULTS AND CONCLUSIONCompared with SAS, SPSS19.0 allowed convenient analysis of categorical repeated measurement data using GEE and GLMMs.
Linear Models ; Models, Statistical ; Software
2.Estimation on gene-environment interaction in the partial case-control study.
Jian-ling BAI ; Peng-cheng XUN ; Yang ZHAO ; Hao YU ; Hong-bing SHEN ; Qing-yi WEI ; Feng CHEN
Chinese Journal of Epidemiology 2006;27(1):72-75
OBJECTIVETo introduce the approaches for estimating gene-environment interaction based on partial case-control studies.
METHODSThe effects of logistic model and log-linear model for estimating the main effects and gene-environment interaction effect were estimated by means of maximum likelihood methods in traditional case-control studies, case-only studies and partial case-control studies, respectively. An example was also illustrated.
RESULTSIn traditional case-control study with complete data, the results of logistic model and log-linear model were equivalent. In case-only study without any information about controls, the logistic model can also efficiently estimate gene-environment interaction. In partial case-control study, environmental information was collected from all of the cases and controls, while genetic information was only collected from cases. For this case-control study with incomplete data, a suitable parameterized log-linear model could simultaneously and efficiently estimate the main effect of environment and gene-environment interaction, whereas the logistic model could not.
CONCLUSIONFor a partial case-control study, log-linear model could estimate not only the main effect of environment but also gene-environment interaction. If genotype and exposure were independent, estimators from partial case-control were as precisely as those from complete-data case-control studies.
Case-Control Studies ; Environment ; Genotype ; Humans ; Linear Models ; Logistic Models ; Models, Statistical ; Reproducibility of Results
3.The impact of incorrectly-measured variables when mixed with precisely measured variables on the study of validity in epidemiological research.
Mei-Xia YANG ; Yi-Biao ZHOU ; Qing-Wu JIANG
Chinese Journal of Epidemiology 2007;28(8):810-813
OBJECTIVETo explore the impact of measurement error on the associated effects under the incorrectly-measured variables when mixed with precisely measured variables.
METHODSBased on the functions of measurement error, correlation of incorrectly-measured predictors and precisely measured explanatory variables, number of precisely measured explanatory variables and associated effect, the 'R Project for Statistical Computing' method is used to analyze the impact of measurement on the validity of a study.
RESULTSUnder the scenario that the continuous response Y and the continuous explanatory Z are precisely measured but the continuous predictor X is incorrectly-measured, when focusing on inference about the effect of X on Y, the non-differential measurement error always makes the value of estimated effect less than the actual value, and the attenuation effect of measurement error more closely worsens the correlation of X and Z. Under a misclassification dichotomous predictor X with an additional precisely measured explanatory variable Z and focusing on inference about the effect of X on Y, the misclassification bias is not only related to the sensitivity and specificity of exposure measurement, but also to the correlation between X and Z and exposure proportion of X. The attenuation factor (AF) decreases gradually with the increasing correlation between X and Z. For instance, in the p = 0.5 scenario, AF is 1.419, and the estimated effect of dichotomous predictor X on continuous response Y is more than the actual effect. When it increases to 0.9, AF is 0.474, the estimated effect becomes less than the true effect.
CONCLUSIONIn the studies of the impact of measurement error in linear regression with additional precisely measured explanatory variables, the impact of measurement error on the associated effect is relatively complex, suggesting that it is necessary to control and to assess the measurement error bias in order to correctly interpret the results of a study.
Bias ; Epidemiologic Research Design ; Epidemiologic Studies ; Linear Models ; Models, Statistical
4.Analysis of periodontal data using mixed effects models.
Journal of Periodontal & Implant Science 2015;45(1):2-7
A fundamental problem in analyzing complex multilevel-structured periodontal data is the violation of independency among the observations, which is an assumption in traditional statistical models (e.g., analysis of variance and ordinary least squares regression). In many cases, aggregation (i.e., mean or sum scores) has been employed to overcome this problem. However, the aggregation approach still exhibits certain limitations, such as a loss of power and detailed information, no cross-level relationship analysis, and the potential for creating an ecological fallacy. In order to handle multilevel-structured data appropriately, mixed effects models have been introduced and employed in dental research using periodontal data. The use of mixed effects models might account for the potential bias due to the violation of the independency assumption as well as provide accurate estimates.
Bias (Epidemiology)
;
Data Interpretation, Statistical
;
Dental Research
;
Least-Squares Analysis
;
Linear Models
;
Models, Statistical
5.A Longitudinal Study of BDNF Promoter Methylation and Depression in Breast Cancer.
Hee Ju KANG ; Jae Min KIM ; Seon Young KIM ; Sung Wan KIM ; Il Seon SHIN ; Hye Ran KIM ; Min Ho PARK ; Myung Geun SHIN ; Jung Han YOON ; Jin Sang YOON
Psychiatry Investigation 2015;12(4):523-531
OBJECTIVE: Brain-derived neurotrophic factor (BDNF) is investigated in depression related to medical disorders and its secretion is influenced by epigenetic factors. We investigated the association between BDNF promoter methylation and depression following mastectomy for breast cancer. METHODS: In total, 309 patients with breast cancer were evaluated 1 week after mastectomy, and 244 (79%) were followed up 1 year later. Depression was diagnosed (major or minor depressive disorder) according to DSM-IV criteria and depression severity was estimated by Montgomery-Asberg Depression Rating Scale (MADRS). We assessed BDNF promoter methylation using leukocyte DNA. The effects of BDNF methylation on depression diagnosis and severity were investigated using multivariate logistic and linear regression models, respectively. The two-way interaction between BDNF methylation and the val66met polymorphism on depression was also evaluated using multivariate logistic regression models. RESULTS: Higher BDNF methylation was independently associated with depression diagnosis and with more severe symptoms at both 1 week and 1 year after mastectomy. No significant methylation-genotype interactions were found. CONCLUSION: A role for BDNF in depression related to breast cancer was supported. Indeed, the association between depression and BDNF methylation may be useful for identifying patients who are at high risk for depression and for suggesting directions for promising drug research.
Brain-Derived Neurotrophic Factor*
;
Breast Neoplasms*
;
Breast*
;
Depression*
;
Diagnosis
;
Diagnostic and Statistical Manual of Mental Disorders
;
DNA
;
DNA Methylation
;
Epigenomics
;
Humans
;
Leukocytes
;
Linear Models
;
Logistic Models
;
Longitudinal Studies*
;
Mastectomy
;
Methylation*
6.An Introduction to Logistic Regression: From Basic Concepts to Interpretation with Particular Attention to Nursing Domain.
Journal of Korean Academy of Nursing 2013;43(2):154-164
PURPOSE: The purpose of this article is twofold: 1) introducing logistic regression (LR), a multivariable method for modeling the relationship between multiple independent variables and a categorical dependent variable, and 2) examining use and reporting of LR in the nursing literature. METHODS: Text books on LR and research articles employing LR as main statistical analysis were reviewed. Twenty-three articles published between 2010 and 2011 in the Journal of Korean Academy of Nursing were analyzed for proper use and reporting of LR models. RESULTS: Logistic regression from basic concepts such as odds, odds ratio, logit transformation and logistic curve, assumption, fitting, reporting and interpreting to cautions were presented. Substantial shortcomings were found in both use of LR and reporting of results. For many studies, sample size was not sufficiently large to call into question the accuracy of the regression model. Additionally, only one study reported validation analysis. CONCLUSION: Nursing researchers need to pay greater attention to guidelines concerning the use and reporting of LR models.
Humans
;
Logistic Models
;
*Models, Statistical
;
Odds Ratio
;
Publishing/standards
;
*Research
7.Developing a Statistical Software for Predicting Hospital Bankruptcy using Data Mining Tool.
Hye Jung CHANG ; Maeng Seok NOH
Journal of Korean Society of Medical Informatics 2001;7(3):9-16
Since the hospital bankruptcy rate is increasing, it has been an important issue to predict the bankruptcy of hospital using the existing hospital management information. Fortunately, the implementation of data mining methodology and decision support system(DSS) are becoming popular. Therefore, this study developed the statistical software for predicting hospital bankruptcy using data mining tool. Stepwise procedures were taken as follows: 1) adopting the HGLM and Logit Models; 2) implementing the input and output processes; 3) linking to the iBITs interface, the data miming tool; and 4) evaluating the software by fitting the hospital management data in practice. The software is written in Visual C++ 5.0 under windows NT/95, and allows the interconnection with other interfaces and libraries. This program initiates encouragement of implementation of DSS models using data mining methodology, in health care fields. This kind of software will play a pivotal role in improving the efficiency and adequacy of managing health care institutions.
Bankruptcy*
;
Data Mining*
;
Delivery of Health Care
;
Logistic Models
;
Models, Statistical
8.Risk Factors Related to Development of Delirium in Hospice Patients.
Hae Jin KO ; Chang Ho YOUN ; Seung Eun CHUNG ; A Sol KIM ; Hyo Min KIM
Korean Journal of Hospice and Palliative Care 2014;17(3):170-178
PURPOSE: Delirium is a common and serious neuropsychiatric complication among terminally ill cancer patients. We investigated risk factors related to the development of delirium among hospice care patients. METHODS: Between May 2011 and September 2012, we included patients who were mentally alert and had no psychiatric disease or drug addiction at the hospice ward of two local hospitals. Among them, participants who had been diagnosed with delirium by two doctors according to the DSM-IV (Diagnostic and Statistical Manual of Mental Disorders-4th edition) criteria were grouped as Delirium Group. We analyzed results of psychometric and other laboratory tests performed at the time of patient's admission - psychometric tests included cognitive function (mini-mental status examination, MMSE), depression (Beck Depression Inventory, BDI), anxiety, and insomnia (Insomnia Severity Index, ISI). Logistic regression analysis was used to compare delirium and the related factors. Cox's proportional hazard model was performed using significant factors of logistic regression analysis. RESULTS: Of the 96 patients who met the inclusion criteria, 41 (42.7%) developed delirium. According to the logistic regression analysis, primary cancer site, cognitive impairment (MMSE<24), depression (BDI> or =16), and insomnia (ISI> or =15) were significant factors related to delirium. Among the four factors, depression (OR 5.130; 95% CI, 2.009~13.097) and cognitive impairment (OR 5.130; 95% CI, 2.009~13.097) were found significant using Cox's proportional hazard model. CONCLUSION: The development of delirium was significantly related to depression and cognitive impairment among patients receiving hospice care. It is necessary to carefully monitor depression and cognitive function in hospice care.
Anxiety
;
Delirium*
;
Depression
;
Diagnostic and Statistical Manual of Mental Disorders
;
Hospice Care
;
Hospices*
;
Humans
;
Logistic Models
;
Proportional Hazards Models
;
Psychometrics
;
Risk Factors*
;
Sleep Initiation and Maintenance Disorders
;
Substance-Related Disorders
;
Terminally Ill
9.Development of Statistical Model for Predicting Prostate Cancer in Patients Requiring Prostate Biopsy.
Taek Woo CHO ; Se Hyun KIM ; Dong PARK
Korean Journal of Urology 2004;45(10):1014-1020
PURPOSE: Patients with an abnormal digital rectal examination(DRE) or elevated serum prostate specific antigen(PSA) level proceed to a transrectal biopsy of the prostate. However, cancer detection is not predictable. There is a need to develop a statistical model for predicting the likelihood of prostate cancer for there to be confidence about the result of a biopsy. MATERIALS AND METHODS: Patients with prostatism were evaluated based upon the recommendation of the International Consultation on benign prostatic hyperplasia(BPH). Amongst the patients evaluated, 141 revealed an abnormal DRE and/or serum PSA. A transrectal ultrasonography(TRUS) and transrectal biopsy was performed in all the patients. 38 of the above were diagnosed with prostate cancer and 103 with BPH or prostatitis. A logistic regression model was used to identify the variables with the most independent influence on prostate cancer and determine the most parsimonious combination of variables for predicting prostate cancer. RESULTS: Age, hematuria, nocturia and a combination of urinary symptoms (incomplete emptying, frequency, urgency and nocturia), DRE, PSA and TRUS-hypoechoic lesion were significant variables for separately predicting prostate cancer. Among these, age, DRE, PSA and TRUS-hypoechoic lesion were independent predictors. The probability of prostate cancer(P) =exp(-9.7770+0.0807xage+1.4079xDRE+0.0257xPSA+1.0904xTRUS- hypoechoic lesion)/{(1+exp(-9.7770+0.0807xage+1.4079xDRE+0.0257xPSA+1.0904xTRUS-hypoechoic lesion)}. CONCLUSIONS: A useful predictive model of prostate cancer has been developed using logistic regression analysis. This model suggests that patients with a high probability(P), but negative biopsy, would require a repeat biopsy. However, a low probability(P), and negative biopsy, would be suggestive of no hidden disease.
Biopsy*
;
Hematuria
;
Humans
;
Logistic Models
;
Models, Statistical*
;
Nocturia
;
Prostate*
;
Prostatic Neoplasms*
;
Prostatism
;
Prostatitis
10.Community Based Study for Stress and It's Related Factors.
Gyung Jae OH ; Jeong Mi LEE ; Sang Sun KIL ; Keun Sang KWON
Korean Journal of Preventive Medicine 2003;36(2):125-130
OBJECTIVES: This study evaluated the stress of community residents using the General Health Questionnaire, GHQ-60, as an instrument of stress measurement. METHODS: The study included 2100 residents, aged 20 and over, living in three areas, a large city, a medium sized city and a rural area, between June and September 2001. A questionnaire interviewing method was used to collect data. The data were analyzed using a t-test, ANOVA, Pearson's correlation coefficients and multiple regression analysis. RESULTS: In this study, the degree of stress, as measured by the GHQ-60, was shown to be significantly higher in the following categories: females, people over 60 years old, people engaged in the primary industries and labor work, low incomes, the divorced and the bereaved, people who received no more than an elementary education, people who suffer from chronic diseases and non-exercisers. A factor analysis suggested that there were three factors of social dysfunction factors; psychosomatic symptom, and depression and anxiety. The social dysfunction factors was statistically significant for the groups described above. The factor of psychosomatic symptoms was statistically significant in the rural residents, and in the groups described above. The depression and anxiety factor was statistically significant in the large city residents, people aged between 20-29 years, students, unmarried persons, university graduates and those having suffered from chronic diseases. From the multiple linear regression analyses, chronic disease, exercise, gender and income, proved to be significant stress related factors. CONCLUSIONS: This study suggests that special attention should be given to the management of the chronic invalided, non-exercisers, females and small income earners, in order to maintain and promote the psychological health of residents in a community.
Anxiety
;
Chronic Disease
;
Depression
;
Divorce
;
Education
;
Factor Analysis, Statistical
;
Female
;
Humans
;
Linear Models
;
Methods
;
Single Person