Common course for all Ph.D. students



Lecturer: associated prof. Ladislav Dusek, Ph.D.
Lesson: credit


The course is aimed at intensive education of Ph.D. students, physicians or other specialists. Participants should learn the basic principles of data analysis, visualization of data, and statistical hypothesis testing. Several specialized lessons provide foundations of multivariate analysis, survival analysis, and predictive modelling of clinical data. Understanding the principles of statistical testing, multivariate analysis, and predictive modelling together with a review of international literature on these topics is the desired output of the course for participants. Application of statistical method is demonstrated in the software Statistica for Windows.
Five afternoon lessons (20 hours in total)

  1. Statistics in medical research - basic principles I. Introduction into the basic principles of statistical data analysis. Concept of probability and its presentation, principles of experiment design, principles of hypothesis testing. Nominal, ordinal and continuous data in clinical research and methods of their visualization. Clinical data "specialties" and consequences for analysis. Data description, variability and data centre quantification, distribution of data. Distribution function and its application for graphical representation of data distribution. Definition of calibration, prognosis, model.
  2. Statistics in medical research - basic principles II. Model distributions and their application (normal, log-normal, alternative, binomial, Poisson, Student, distribution of t, F, and c2 statistics). Confidence interval estimation, presentation of arithmetic/geometric mean, variability, and estimates of other parameters. Estimate of median. Summary statistics of continuous and discrete data. Examples of summary statistics presentation.
  3. Statistics in medical research - basic principles III. Data preparation. Graphical tools for data visualization - explorative analysis /"PP plots, QQ plots, normal probability plots, box-and-whisker plots, scatter plots, stem and leaf display, histograms, 3D histograms, matrix plots - face plots, contour plots, surface plots"/. Data transformations in analytical practice. Identification of outliers. Use and misuse of computers in clinical data analysis. Non-parametric methods - alternative for data which does not fit prerequisites of parametric techniques. Examples of non-parametric techniques. Examples summarizing lessons I - III.
  4. Univariate statistics - continuous data. Univariate analysis of continuous data. One-sample and two-sample tests. T-test for independent and dependent (paired) data. Analysis of variance (ANOVA) - basic principles of one-way and multi-way ANOVA, tests of contrasts. Non-parametric methods (Mann-Whitney test, Wald-Worowitz test, Kolmogorov-Smirnov two-sample test, Kruskal-Wallis test). Graphical methods for visualization of results of the above-mentioned tests.
  5. Univariate statistics - discrete data. Univariate analysis of discrete data. One-sample and two-sample tests. Presentation of percentages and estimates of parameters of data expressed as percentages. Binomial test. Fisher exact test. Goodness of fits test and its application to clinical data. Frequency table analysis - other tests.
  6. Basics of correlation and regression. Principles of correlation analysis. Parametric and non-parametric correlation. Principles of regression analysis. Linear model and its analysis. Application and graphical presentation of correlation and regression. Examples and basics of polynomial and non-linear regression.
  7. Principles of multivariate and logistic regression. Multivariate and logistic regression - predictive methods for clinical data. Principles of multivariate regression. Quality of models and possibility of errors. Application and examples of multivariate regression for prediction of practically important clinical parameters. Logistic regression models - a possible tool for individual prediction of patients. Presentation of predictive models. Examples.
  8. Survival analysis. Probability of survival. Kaplan-Meier survival analysis and parameter estimates /median survival times …/. Range of approaches for comparison of two or more survival curves /Log-rank test, hazard ratio, log rank for trends, confidence intervals for survival probability/. "Cohort life tables" and their analysis of survival. Modelling of survival, Cox regression. Examples and application. Design of studies focused on survival analysis - quantitative aspects of experimental design, sample size estimation. Survival analysis for stratified clinical trials. EORTC standards for experimental design of survival analysis. Internet and survival analysis: consultation on trials aimed at survival analysis, software for survival analysis. Nomograms for design of survival analysis trials.
  9. Multivariate analysis of clinical data; introduction into modern methods of analysis of huge data. Principles of multivariate methods and their application to clinical data analysis. Multivariate and univariate data analysis - mutual collaboration or discrepancy? Multivariate data exploration, available tests for multivariate distribution. Multivariate similarity/distance of objects or variables - review of important metrics. Dynamic regression models. Neural networks as possible modelling technique. Data mining and automated analysis of data. Optimalization of experiments; application of multivariate methods in sampling design.