William C. Rinaman

Introduction to SPSS

This chapter introduces several basic SPSS procedures that are used in the analysis of a data set. The chapter explains the structure of SPSS data files, how to open an SPSS data file, and how to import into SPSS data contained in an Excel file. The chapter also explains how to select cases for an analysis, display variables listed in dialog boxes in alphabetical order, label and print output, paste output into a Microsoft Word document, and save data and output as SPSS or Excel files.

Statistical literacy for clinical practitioners

Overview.- The Evidence Pyramid.- Case Study.- Case-control Study.- Randomized Controlled Trial.- Meta-analysis.- References.- Exercise Questions.

Introduction to Statistical Inference

This is an introduction to the two key tools for statistical inference. Confidence intervals on a population mean are introduced. This is followed by an introduction to the ideas behind hypothesis testing. They are applied to test a population mean. Since these procedures can depend, in the case of small samples, on the population distribution being normal, tests of this assumption are discussed. When the usual tests cannot be used, the Wilcoxon signed ranks test is introduced. This is followed by a discussion of statistical power and the difference between clinical significance and statistical significance.

Comparing Means of Related Samples

This chapter reviews the paired-samples t-test and the repeated measures analysis of variance (ANOVA). These are inferential statistics commonly used to test the difference between the means of populations that are related to each other, such as the means of a quantitative measurement taken of the same group of participants on two or more occasions. Because the ANOVA assumes the presence of a condition known as sphericity, the chapter also reviews Mauchly’s test of sphericity and methods for hypothesis testing when sphericity cannot be assumed.

Comparing Means of Independent Samples

This chapter reviews the independent-samples t-test and the one-way analysis of variance, inferential statistics that are commonly used to test null and alternative hypotheses about mean differences among independent populations. Because both procedures assume equal population variances, Levene’s test for homogeneity of variances is discussed, as are methods for hypothesis testing when homogeneity of variances cannot be safely assumed. The chapter continues by using a measure of effect size, partial eta squared, to distinguish between statistical and clinical significance, and concludes with a discussion of post hoc multiple comparisons and contrast analysis.

Regression Analysis of Count Data

This chapter reviews negative binomial regression. Often used to document incidence and mortality rates, this form of regression generates a rate ratio to assess the degree of relationship between a predictor variable and the frequency with which an event occurs over a given period of time. The chapter begins with a discussion of the case of a single predictor variable, and then moves on to a discussion of two or more predictors, and of testing for the presence of interactions. As an example of the difference between cumulative incidence and incidence rate, the concept of person-years, and the use of an offset variable, the chapter concludes with an application of negative binomial regression to count data collected over unequal follow-up times.

Describing the Distribution of a Quantitative Variable

This chapter reviews measures of central tendency and spread, and graphical techniques that are commonly used to describe the distributions of quantitative data. Included are the arithmetic mean and median; interquartile range, variance and standard deviation; skewness, kurtosis, and outliers; and histograms, stem-and-leaf plots, box plots, and clustered bar charts. The standard error of the mean and the 95 % confidence interval are described briefly. The chapter concludes with a discussion of transformations and the geometric mean.

Multiple Linear Regression

This chapter provides an overview of multiple linear regression, a statistical technique that predicts values of a quantitative dependent variable from values of two or more independent variables. By including more than one independent variable, a multiple linear regression can often account for more variability in the dependent variable than can a simple regression, can assess the relationship between the dependent variable and an independent variable after controlling for the presence of other independent variables, and can determine whether the effect of an independent variable varies across levels of another. Topics reviewed include the multiple correlation coefficient, adjusted R 2,interpreting and testing unstandardized and standardized slope coefficients, using categorical and dummy variables as predictors, and testing for the presence of interaction effects.

Relationships in Categorical Data

This chapter investigates relationships in categorical data. It begins with a discussion of contingency tables and clustered bar charts as descriptive measures. The chi-square test for contingency tables is discussed. If the two categorical variables are found to be related, then the strength of that relationship is measured using Cramer’s V for nominal variables and Gamma for ordinal variables.

Relationships in Quantitative Data

This chapter investigates assessing relationships between two quantitative variables. Scatter plots are introduced as a graphical way to determine whether a relationship exists between the two variables and assess the shape, direction, and strength of the relationship. When the relationship is linear, the Pearson correlation coefficient is introduced to measure the strength of the relationship. Tests and confidence intervals on the Pearson correlation coefficient are discussed. For nonlinear relationships Spearman’s rho is discussed.