The gamma generalized linear model, log transformation, and the robust Yuen-Welch test for analyzing group means with skewed and heteroscedastic data

ABSTRACT Alternatives for positively skewed and heteroscedastic data include the Yuen-Welch (YW) test, data transformations, and the generalized linear model (GzLM). Because the GzLM is rarely considered in psychology compared to the other two, we compared these strategies conceptually and empirically. The YW test generally has satisfactory power, but its trimmed mean can deviate substantially from the arithmetic mean, which is often the desired parameter. The gamma GzLM can be used as a substitute for the log transformation and addresses the limitations in inference for the YW and data transformations.

[1]  J. Copas,et al.  Using generalized linear models (GLMs) to model errors in motor performance. , 1991, Journal of motor behavior.

[2]  D. W. Zimmerman.  A note on preliminary tests of equality of variances. , 2004, The British journal of mathematical and statistical psychology.

[3]  Robert A Cribbie,et al.  ANOVA and the variance homogeneity assumption: Exploring a better gatekeeper , 2018, The British journal of mathematical and statistical psychology.

[4]  Andrew F Hayes,et al.  Further evaluating the conditional decision rule for comparing two independent means. , 2007, The British journal of mathematical and statistical psychology.

[5]  S. Shapiro,et al.  An Analysis of Variance Test for Normality (Complete Samples) , 1965 .

[6]  Equivalence of Population Variances: Synchronizing the Objective and Analysis , 2018 .

[7]  J. Gastwirth,et al.  The impact of Levene’s test of equality of variances on statistical theory and practice , 2009, 1010.0308.

[8]  Eric R. Ziegel,et al.  Generalized Linear Models , 2002, Technometrics.

[9]  Andrew M. Jones.  The Elgar Companion to Health Economics , 2007 .

[10]  E. Chen.  The Power of the Shapiro-Wilk W Test for Normality in Samples from Contaminated Normal Distributions , 1971 .

[11]  Y. B. Wah,et al.  Power comparisons of Shapiro-Wilk , Kolmogorov-Smirnov , Lilliefors and Anderson-Darling tests , 2011 .

[12]  M C Hornbrook,et al.  Modeling risk using generalized linear models. , 1999, Journal of health economics.

[13]  E. Ziegel.  Introduction to Robust Estimation and Hypothesis Testing (2nd ed.) , 2005 .

[14]  W. Manning,et al.  Estimating Log Models: To Transform or Not to Transform? , 1999, Journal of health economics.

[15]  W. R. Schucany,et al.  Preliminary Goodness-of-Fit Tests for Normality do not Validate the One-Sample Student t , 2006 .

[16]  R. Ratcliff.  Methods for dealing with reaction time outliers. , 1993, Psychological bulletin.

[17]  Douglas G Bonett,et al.  Statistical inference for a linear function of medians: confidence intervals, hypothesis testing, and sample size requirements. , 2002, Psychological methods.

[18]  P. Younger.  The Elgar Companion to Health Economics , 2006 .

[19]  K. Yuen,et al.  The two-sample trimmed t for unequal population variances , 1974 .

[20]  David M Erceg-Hurn,et al.  Modern robust statistical methods: an easy way to maximize the accuracy and power of your research. , 2008, The American psychologist.

[21]  N. Balakrishna,et al.  Communications in Statistics-Theory and Methods , 2012 .

[22]  John Fox,et al.  Applied Regression Analysis and Generalized Linear Models , 2008 .

[23]  H. Levene.  Robust tests for equality of variances , 1961 .

[24]  Morton B. Brown,et al.  Robust Tests for the Equality of Variances , 1974 .

[25]  N. Duan.  Smearing Estimate: A Nonparametric Retransformation Method , 1983 .

[26]  Miguel A. García-Pérez,et al.  Statistical Conclusion Validity: Some Common Threats and Simple Remedies , 2012, Front. Psychology.

[27]  J. Simons,et al.  Inference in Regression Models of Heavily Skewed Alcohol Use Data: A Comparison of Ordinary Least Squares, Generalized Linear Models, and Bootstrap Resampling , 2007, Psychology of addictive behaviors : journal of the Society of Psychologists in Addictive Behaviors.

[28]  R Core Team,et al.  R: A language and environment for statistical computing. , 2014 .

[29]  M. Angermeyer,et al.  A comparison of methods to handle skew distributed cost variables in the analysis of the resource consumption in schizophrenia treatment. , 2002, The journal of mental health policy and economics.

[30]  W. Manning,et al.  The logged dependent variable, heteroscedasticity, and the retransformation problem. , 1998, Journal of health economics.

[31]  R. Grissom,et al.  Heterogeneity of variance in clinical data. , 2000, Journal of consulting and clinical psychology.

[32]  Rand R. Wilcox,et al.  Trimming, Transforming Statistics, And Bootstrapping: Circumventing the Biasing Effects Of Heterescedasticity And Nonnormality , 2002 .

[33]  S. Shapiro,et al.  An analysis of variance test for normality ( complete samp 1 es ) t , 2007 .

[34]  R. Cribbie,et al.  The expanding role of quantitative methodologists in advancing psychology. , 2009 .

[35]  D. Cox,et al.  An Analysis of Transformations , 1964 .

[36]  T. Micceri.  The unicorn, the normal curve, and other improbable creatures. , 1989 .

[37]  Julia Kastner,et al.  Introduction to Robust Estimation and Hypothesis Testing , 2005 .

[38]  J. Fox,et al.  Applied Regression Analysis and Generalized Linear Models , 2008 .