Statistical methods in medical research
-
Stat Methods Med Res · Dec 2006
Correlation coefficients in medical research: from product moment correlation to the odds ratio.
Presentation of effect sizes that can be interpreted in terms of clinical or practical significance is currently urged whenever statistical significance (a 'p-value') is reported in research journals. However, which effect size and how to interpret it are not yet clearly delineated. The present focus is on effect sizes indicating strength of correlation, that is, effect sizes that describe the strength of monotonic association between two random variables X and Y in a population. ⋯ Suggestions are made for the future use of measures of association in research to facilitate considerations of clinical significance, emphasizing distribution-free effect sizes such as the Spearman correlation coefficient and Kendall's coefficient of concordance for ordinal versus ordinal associations, weighted and intraclass kappa for binary versus binary associations and risk difference (RD) for binary versus ordinal association.
-
Publication bias has been previously identified as a threat to the validity of a meta-analysis. Recently, new evidence has documented an additional threat to validity, the selective reporting of trial outcomes within published studies. ⋯ Some articles might report only a selection of those outcomes, perhaps those with statistically significant results. In this article, we review this problem while addressing the questions: what is within-study selective reporting? how common is it? why is it done? how can it mislead? how can it be detected?, and finally, what is the solution? We recommend that both publication bias and selective reporting should be routinely investigated in systematic reviews.
-
Stat Methods Med Res · Feb 2004
ReviewMixed effects multivariate adaptive splines model for the analysis of longitudinal and growth curve data.
In this article, I review the use of nonparametric methods in the analysis of longitudinal and growth curve data, particularly the multivariate adaptive splines models for the analysis of longitudinal data (MASAL). These methods combine nonparametric techniques (B-splines, kernel smoothing, piecewise polynomials) and models with random effects, and provide fruitful alternatives to mixed effects linear models. ⋯ The analysis of a real example is also presented to illustrate the application and interpretation of MASAL. Open questions are posed for further investigation.
-
Stat Methods Med Res · Feb 2004
ReviewFunctional data analysis in longitudinal settings using smoothing splines.
Data in many experiments arise as curves and therefore it is natural to use a curve as a basic unit in the analysis, which is termed functional data analysis (FDA). In longitudinal studies, recent developments in FDA have extended classical linear models and linear mixed effects models to functional linear models (also termed varying-coefficient models) and functional mixed effects models. ⋯ Due to the connection between smoothing splines and linear mixed effects models, functional mixed effects models can be fitted using existing software such as SAS Proc Mixed. A case study is presented as an illustration.