基于logistic回归的连续型解释变量离散化方法效果比较

doi:10.3969/j.issn.1006-5253.2022.04.016

摘要
图/表
参考文献
相关文章 (15)

全文: PDF (3617 KB) HTML (1 KB)
输出: BibTeX | EndNote (RIS)

摘要目的探讨当连续型自变量和logit^π不满足线性关系时，不同数据离散化方法的优劣，为logistic回归分析中连续型自变量离散化处理提供参考。方法运用R软件，采用病例对照研究从效应量、自变量个数、样本量3个角度产生模拟数据，进一步对连续型自变量采用不同离散化处理后拟合logistic回归模型，比较不同方法拟合效果。结果不同模拟数据集研究结果均显示，4种连续型变量处理方法拟合logistic回归模型时，“OR 值最大化分类法”能更好地筛选出对结局变量有影响的自变量，并且其模型拟合效果亦最好，表现为其 AIC 最小，Nagelkerke R²最大，正确合计率也较高。结论当连续型自变量与logit ^π 之间呈非单调变化关系时，在拟合logistic回归模型时建议使用双界点 OR 值最大法对数据进行离散化。

	服务

	把本文推荐给朋友
	加入我的书架
	加入引用管理器
	E-mail Alert
	RSS
	作者相关文章
	何贤英

关键词 ： logistic回归, 半参数回归, 离散化, OR值最大法

Abstract：Objective To explore the advantages and disadvantages of different data discretization methods when the continuous independent variable and logit-π-do not satisfy the linear relationship, and to provide reference for the continuous independent variable discretization in logistic regression analysis.Methods A casecontrol study was used to generate simulated data from the three perspectives of effect size, number of independent variables, and sample size with R software. The continuous independent variables were further processed by different discretization to fit logistic regression models, and the fitting effects of different methods were compared.Results The results of different simulation data sets showed that when the four continuous variable processing methods were used to fit the logistic regression model, "maximum OR values method" could better screen out meaningful influencing factors. At the same time, its model fitting effect was also the best, which was shown as the smallest-AIC, the largest Nagelkerke R², and the higher correct total rate.Conclusion It is recommended to use "maximum-OR-values method" to discrete continuous variables if the relationship between these variables and logit^π is non-monotonic.

Key words： logistic regression semi-parametric regression discretization maximum-OR-value method

收稿日期: 2022-05-14

通讯作者: 何贤英，Email：hexianying@zzu.edu.cn

引用本文:

何贤英. 基于logistic回归的连续型解释变量离散化方法效果比较[J]. 中国医院统计, 2022, 29(4): 312-317.
He Xianying. Comparison of the effects of discretization methods of continuous explanatory variables based on logistic regression. journal1, 2022, 29(4): 312-317.

链接本文:

http://manu58.magtech.com.cn/Jwk_zgyytj/CN/10.3969/j.issn.1006-5253.2022.04.016 或 http://manu58.magtech.com.cn/Jwk_zgyytj/CN/Y2022/V29/I4/312

［1］SCHULGEN G, LAUSEN B, OLSEN J H, et al. Outcome-oriented cutpoints in analysis of quantitative exposures［J］. Am J Epidemiol, 1994, 140(2):172-184. DOI:10.1093/oxfordjournals.aje.a117227.
［2］ABDOLELL M, LEBLANC M, STEPHENS D, et al. Binary partitioning for continuous longitudinal data: Categorizing a prognostic variable［J］. Stat Med, 2002, 21(22):3395-3409. DOI:10.1002/sim.1266.
［3］ROSNOW R L, ROSENTHAL R, RUBIN D B. Contrasts and correlations in effect-size estimation［J］. Psychol Sci, 2000, 11(6):446-453. DOI:10.1111/1467-9280.00287.
［4］SINK C A, MVUDUDU N H. Statistical power, sampling, and effect sizes［J］. Couns Outcome Res Eval, 2010, 1(2):1-18. DOI:10.1177/2150137810373613.
［5］何贤英，赵志，黄嘉玲，等．自变量连续型测定值及基于中位数的0～1转化值拟合logistic回归模型的效果比较［J］．中国卫生统计，2017，34(6)：869-872．
［6］WAND M P, COULL B A, FRENCH J L, et al. SemiPar 1.0. R package［A］. Vienna: Comprehensive R Archive Network Project, 2005.
［7］何贤英，赵志，温兴煊，等．logistic回归中连续型自变量离散化为二分类变量时适宜分界点的确定［J］．中国卫生统计，2015，32(2)：275-277．
［8］WILLIAMS B A, MANDREKAR J N, MANDREKAR S J, et al. Finding optimal cut points for continuous covariates with binary and time-to-event outcomes［R/OL］.Minnesota:Department of Health Sciences Research Mayo Clinic Rochester,2006.https://www.researchgate.net/publication/241590524_Finding_Optimal_Cutpoints_for_Continuous_Covariates_with_Binary_and_Time-to-Event_Outcomes.
［9］ALTMAN D G, LAUSEN B, SAUERBREI W, et al. Dangers of using “optimal” cutpoints in the evaluation of prognostic factors［J］. J Natl Cancer Inst, 1994, 86(11):829-835. DOI:10.1093/jnci/86.11.829.
［10］FARAGGI D, SIMON R. A simulation study of cross-validation for selecting an optimal cutpoint in univariate survival analysis［J］. Stat Med, 1996, 15(20):2203-2213. DOI:10.1002/(SICI)1097-0258(19961030)15:20<2203:AID-SIM357>3.0.CO;2-G.
［11］HILSENBECK S G, CLARK G M. Practical p-value adjustment for optimally selected cutpoints［J］. Stat Med, 1996, 15(1):103-112. DOI:10.1002/(SICI)1097-0258(19960115)15:1<103:AID-SIM156>3.0.CO;2-Y.
［12］CONTAL C, O′QUIGLEY J. An application of changepoint methods in studying the effect of age on survival in breast cancer［J］. Comput Stat Data Anal, 1999, 30(3):253-270. DOI:10.1016/S0167-9473(98)00096-6.
［13］MAZUMDAR M,GLASSMAN J R.Categorizing a prognostic variable:Review of methods, code for easy implementation and applications to decision-making about cancer treatments［J］.Stat Med,2000,19(1):113-132.DOI:10.1002/(sici)1097-0258(20000115)19:1<113:aid-sim245>3.0.co;2-o.
［14］MAZUMDAR M, SMITH A, BACIK J. Methods for categorizing a prognostic variable in a multivariable setting［J］. Stat Med, 2003, 22(4):559-571. DOI:10.1002/sim.1333.
［15］CUMSILLE F, BANGDIWALA S I, SEN P K, et al. Effect of dichotomizinlg a continuous variable on the model structure in multiple linear regression models［J］. Commun Stat Theory Methods, 2000, 29(3):643-654. DOI:10.1080/03610920008832507.
［16］ROYSTON P, SAUERBREI W, ALTMAN D G. Modeling the effects of continuous risk factors［J］. J Clin Epidemiol, 2000, 53(2):219-221. DOI:10.1016/s0895-4356(99)00163-8.
［17］HEINZL H, TEMPFER C. A cautionary note on segmenting a cyclical covariate by minimum P-value search［J］. Comput Stat Data Anal, 2001, 35(4):451-461. DOI:10.1016/S0167-9473(00)00023-2.
［18］LIQUET B, COMMENGES D. Correction of the P-value after multiple coding of an explanatory variable in logistic regression［J］. Stat Med, 2001, 20(19):2815-2826. DOI:10.1002/sim.916.
［19］CHEN Y M, HUANG J L, HE X Y, et al. A novel approach to determine two optimal cut-points of a continuous predictor with a U-shaped relationship to hazard ratio in survival data: Simulation and application［J］. BMC Med Res Methodol, 2019, 19(1):96. DOI:10.1186/s12874-019-0738-4.
［20］MACCALLUM R C, ZHANG S B, PREACHER K J, et al. On the practice of dichotomization of quantitative variables［J］. Psychol Methods, 2002, 7(1):19-40. DOI:10.1037/1082-989x.7.1.19.
［21］冯国双，陈景武，周春莲．logistic回归应用中容易忽视的几个问题［J］．中华流行病学杂志，2004，25(6)：544-545．DOI:10.3760/j.issn:0254-6450.2004.06.022.
［22］Ruppert D, Wand M P, Carroll R J. Semiparametric regression［M］. New York:Cambridge University Press, 2003.