Table 3 Categories of statistical methods used to assess the statistical content of articles
From: Systematic review of analytical methods applied to longitudinal studies of malaria
Category | Brief description | Include |
|---|---|---|
No statistical methods or descriptive statistics only | Describe basic features of data to provide simple measures of summaries | No statistical content, or descriptive statistics only e.g., percentages, means, standard deviations, standard errors, histograms |
Contingency tables | Cross tabulations used to summarize the relationship between categorical data | Chi-square test, Fisher’s exact test, McNemar’s test |
Epidemiologic statistics | Measures of association for outcome of interest such as disease and some exposure(s) | Relative risk, risk ratios, rate ratios, risk difference, rate difference, odds ratio, log odds, risk difference, attributable risk fraction, sensitivity, specificity |
Multiway tables | Extend two-way relationships to include three or more variables | Mantel–Haenszel procedure, log-linear models, logistic regression |
t-test | Assess mean differences between groups | One-sample, matched-pair, two-sample t-tests |
Pearson’s correlation | Measures linear correlation between two variables | Classical product-moment correlation |
Simple linear regression | Regression that summarizes relationships between two continuous variables, an explanatory and a response | Least-squares regression with one predictor and one response variable |
Multiple regression | Extends the simple regression to include two or more explanatory variables for a response | Polynomial regression and stepwise regression |
Analysis of variance | Assess within and between group differences in means | Analysis of variance, Analysis of covariance, simple linear contrasts, F-tests |
Multiple comparisons | Methods for handling multiple inferences on same data sets | Bonferroni techniques, Scheffé’s contrasts, Duncan multiple-range methods, Newman–Keuls procedure |
Non-parametric test | Tests used when data is not assumed to follow a particular distribution, and are based on ranks of data | Sign test, Wilcoxon signed-rank test, Mann–Whitney test, Kruskal–Wallis test, Friedman test, Kolmogorov–Smirnov test |
Non-parametric correlation | Measure strength and direction of association between two variables | Spearman’s rho, Kendall’s tau, monotone regression, test for trend |
Survival analysis | Methods where outcome variable is the time until the occurrence of an event | Actuarial life table, Kaplan–Meier estimator for survival, survival function, Cox model, other parametric survival models, rate adjustment, log-rank test, Breslow’s test |
Sensitivity analysis | Examines sensitivity of outcome to small changes in parameters of model or in other assumptions | Sample size, multiple outcomes, model distribution assumptions |
Transformation | Use of data transformation often in regression | Natural logarithm, square, cubic |
Cluster analysis | Involves dividing a multivariate dataset into “natural” clusters (groups) for in-depth assessment | Hierarchical, K-means, two-step clustering |
Repeated-measures analysis | Approaches that account for correlation for within-participant observations and non-constant variance in response over time | Generalized estimating equations (GEE), mixed-effects models, repeated measures ANOVA |
Other | Other methods not specified in above | Receiver-operating characteristic, principal component analysis, power analysis, propensity score |