The Comparison of Alternative Smoothing Methods for Fitting Non-Linear Exposure-Response Relationships with Cox Models in a Simulation Study

被引:97
作者
Govindarajulu, Usha S. [1 ]
Malloy, Elizabeth J. [2 ]
Ganguli, Bhaswati [3 ]
Spiegelman, Donna [4 ]
Eisen, Ellen A. [5 ]
机构
[1] Harvard Univ, Sch Med, Cambridge, MA 02138 USA
[2] American Univ, Washington, DC 20016 USA
[3] Univ Calcutta, Kolkata 700073, W Bengal, India
[4] Harvard Univ, Sch Publ Hlth, Cambridge, MA 02138 USA
[5] Univ Calif Berkeley, Berkeley, CA 94720 USA
关键词
penalized spline; simulation; restricted cubic spline; natural spline; fractional polynomial; Cox model; CUBIC SPLINE FUNCTIONS; LUNG-CANCER MORTALITY; REGRESSION-MODELS; TIME; EPIDEMIOLOGY;
D O I
10.2202/1557-4679.1104
中图分类号
Q [生物科学];
学科分类号
07 ; 0710 ; 09 ;
摘要
We examined the behavior of alternative smoothing methods for modeling environmental epidemiology data. Model fit can only be examined when the true exposure-response curve is known and so we used simulation studies to examine the performance of penalized splines (P-splines), restricted cubic splines (RCS), natural splines (NS), and fractional polynomials (FP). Survival data were generated under six plausible exposure-response scenarios with a right skewed exposure distribution, typical of environmental exposures. Cox models with each spline or FP were fit to simulated datasets. The best models, e. g. degrees of freedom, were selected using default criteria for each method. The root mean-square error (rMSE) and area difference were computed to assess model fit and bias (difference between the observed and true curves). The test for linearity was a measure of sensitivity and the test of the null was an assessment of statistical power. No one method performed best according to all four measures of performance, however, all methods performed reasonably well. The model fit was best for P-splines for almost all true positive scenarios, although fractional polynomials and RCS were least biased, on average.
引用
收藏
页数:21
相关论文
共 27 条
[1]   NEW LOOK AT STATISTICAL-MODEL IDENTIFICATION [J].
AKAIKE, H .
IEEE TRANSACTIONS ON AUTOMATIC CONTROL, 1974, AC19 (06) :716-723
[2]  
[Anonymous], 2003, Modelling Survival Data in Medical Research
[3]  
ARMITAGE P, 1961, STOCHASTIC MODELS CA
[4]  
Breslow N. E., 1987, STAT METHODS CANC RE, V2
[5]   Dose-response associations of silica with nonmalignant respiratory disease and lung cancer mortality in the diatomaceous earth industry [J].
Checkoway, H ;
Heyer, NJ ;
Seixas, NS ;
Welp, EAE ;
Demers, PA ;
Hughes, JM ;
Weill, H .
AMERICAN JOURNAL OF EPIDEMIOLOGY, 1997, 145 (08) :680-688
[6]  
COX, 1985, ANAL SURVIVAL DATA, P19
[7]   FLEXIBLE REGRESSION-MODELS WITH CUBIC-SPLINES [J].
DURRLEMAN, S ;
SIMON, R .
STATISTICS IN MEDICINE, 1989, 8 (05) :551-561
[8]   APPROXIMATE REGRESSION-MODELS AND SPLINES [J].
EUBANK, RL .
COMMUNICATIONS IN STATISTICS-THEORY AND METHODS, 1984, 13 (04) :433-484
[9]   FITTING SEGMENTED POLYNOMIAL REGRESSION MODELS WHOSE JOIN POINTS HAVE TO BE ESTIMATED [J].
GALLANT, AR ;
FULLER, WA .
JOURNAL OF THE AMERICAN STATISTICAL ASSOCIATION, 1973, 68 (341) :144-147
[10]   Comparing smoothing techniques in Cox models for exposure-response relationships [J].
Govindarajulu, Usha S. ;
Spiegelman, Donna ;
Thurston, Sally W. ;
Ganguli, Bhaswati ;
Eisen, Ellen A. .
STATISTICS IN MEDICINE, 2007, 26 (20) :3735-3752