Scaling regression inputs by dividing by two standard deviations

被引:1829
作者
Gelman, Andrew [1 ,2 ]
机构
[1] Columbia Univ, Dept Stat, New York, NY 10027 USA
[2] Columbia Univ, Dept Polit Sci, New York, NY USA
关键词
generalized linear models; linear regression; logistic regression; standardization; z-score;
D O I
10.1002/sim.3107
中图分类号
Q [生物科学];
学科分类号
07 [理学]; 0710 [生物学]; 09 [农学];
摘要
Interpretation of regression coefficients is sensitive to the scale of the inputs. One method often used to place input variables on a common scale is to divide each numeric variable by its standard deviation. Here we propose dividing each numeric variable by two times its standard deviation, so that the generic comparison is with inputs equal to the mean +/- 1 standard deviation. The resulting coefficients are then directly comparable for untransformed binary predictors. We have implemented the procedure as a function in R. We illustrate the method with two simple analyses that are typical of applied modeling: a linear regression of data from the National Election Study and a multilevel logistic regression of data on the prevalence of rodents in New York City apartments. We recommend our resealing as a default option-an improvement upon the usual approach of including variables in whatever way they are coded in the data file-so that the magnitudes of coefficients can be directly compared as a matter of routine statistical practice. Copyright (C) 2007 John Wiley & Sons, Ltd.
引用
收藏
页码:2865 / 2873
页数:9
相关论文
共 18 条
[1]
Taxes, cigarette consumption, and smoking intensity [J].
Adda, Jerome ;
Cornaglia, Francesca .
AMERICAN ECONOMIC REVIEW, 2006, 96 (04) :1013-1028
[2]
[Anonymous], 2006, POLARIZED AM DANCE P
[3]
EVALUATING THE RELATIVE IMPORTANCE OF VARIABLES [J].
BLALOCK, HM .
AMERICAN SOCIOLOGICAL REVIEW, 1961, 26 (06) :866-874
[4]
HOW TO STANDARDIZE REGRESSION-COEFFICIENTS [J].
BRING, J .
AMERICAN STATISTICIAN, 1994, 48 (03) :209-213
[5]
Gelman A., 2017, Data Analysis Using Regression and Multilevel/Hierarchical Models
[6]
GELMAN A, 2007, Q J POLITIC IN PRESS
[7]
GELMAN A, 2007, SOCIOLOGICA IN PRESS
[8]
THE FALLACY OF EMPLOYING STANDARDIZED REGRESSION-COEFFICIENTS AND CORRELATIONS AS MEASURES OF EFFECT [J].
GREENLAND, S ;
SCHLESSELMAN, JJ ;
CRIQUI, MH .
AMERICAN JOURNAL OF EPIDEMIOLOGY, 1986, 123 (02) :203-208
[9]
HARRELL FE, 2001, REGRESSION MODELING
[10]
HASTIE TJ, 1990, GEN ADDITIVE