Expertise & Capabilities

Data Visualization and Storytelling (1.0 days):

This course will provide principals to effectively communicate your technical results professionally. The methods will allow you to interactively discover deeper relationships graphically more efficiently. We will provide the foundations for creating better graphical information to accelerate the insight discovery process and enhance the understandability of reported results. First principles and the “human as part of the system” aspects of information visualization from multiple leading sources are explored using representative univariate, multivariate, time series, geographic, and unstructured text data sets.

Data Visualization and Storytelling Objectives:

• Understand data visualization principals
• Know the characteristics of data that help guide appropriate visualizations
• Be able to apply a disciplined process for designing good graphics and dashboards
• Create graphs for different data types to include univariate, multivariate, categorical, text, time series, geographic, and modeling output.
• Be able to tell the story of applied statistical analysis integrating the narrative with appropriate displays

Introduction
• Introduction, definitions, and historical perspectives
• Examples of data visualizations and interactivity
• Characteristics of data

Designing Effective Graphics
• Human side of data visualization
• Principles of good graphic design
• Data visualization methodology and best practices
• Dashboards

Creating the Right Graph
• Univariate plots: distributions, histograms, and boxplots
• Multivariate plots: scatterplots, contour plots, treemaps
• Categorical data plots: pie/doughnut graphs, mosaic plots
• Text/Unstructured data: word clouds, topic plots, word and documents clusters
• Time series data: trend lines/forecasts, waterfall plots, sparklines
• Geographic data: maps, custom maps

Graphics from Modeling
• Model adequacy checks
• Factor profilers and optimization
• Comparing competing models

Telling the story
• Data storytelling steps and principals
• Best practices
• Oral presentation tips

Applied Statistics for Scientists (2.0 days):

This course provides instruction on how to apply scientific approaches to data analysis by focusing on appropriate methods for analysis: descriptive statistics, hypothesis testing, analysis of variance and model building.

Applied Statistics for Scientists Objectives:

• describe and analyze the distribution of data
• develop summary statistics
• generate and analyze statistical intervals and hypothesis tests to make data-driven decisions
• understand issues related to sampling and calculate appropriate sample sizes
• describe the relationship between and among two or more factors or responses

Outline:

1. Descriptive Statistics
• introducing summary statistics
• describing a distribution of values
• creating a summary table and tabulating data

2. Intervals
• producing confidence intervals
• producing tolerance intervals

3. Hypothesis Testing
• introducing hypothesis testing
• performing means (and variance) tests
• performing proportion tests

4. ANOVA
• defining analysis of variance and other terminology
• discussing assumptions and interpretation
• interpreting hypothesis statements for ANOVA
• performing one-way ANOVA

5. Simple Linear Regression
• producing scatterplots and performing correlation
• performing simple linear regression

6. Model Building
• introducing model building
• performing N-way ANOVA
• performing multiple linear regression
• performing ANCOVA

7. Model Diagnostics
• evaluating the validity of model assumptions
• evaluating model fit

Demonstration of Comparability for Chemistry, Manufacturing, and Controls (CMC) in the Pharmaceutical Industry (0.5 days): The FDA comparability guidance (1996) recognizes the need for manufacturers to improve manufacturing processes and analytical procedures without performing additional clinical studies to demonstrate product safety and efficacy. This guidance was extended in ICH Q5E (2004) to provide additional direction for comparing pre- and post-change manufacturing processes.

Across the regulatory documents, there are only high-level recommendations for the design of an equivalence/comparability study and for setting acceptance criteria to assess the impact of the change. These documents do not generally contain prescriptive rules for setting acceptance criteria, study design, or statistical procedures for analysis. One notable exception is the draft guidance published by FDA (2019) concerning demonstration of analytical similarity. EMA also produced a report based on comments offered on a reflection paper concerning analytical comparability and similarity (2018). This course presents statistical approaches for demonstrating comparability consistent with these regulatory guidelines. In addition, this course provides JMP scripts for power and sample size determinations as well as appropriate constants for intervals used to calculate quality ranges:

Demonstration of Comparability for CMC in the Pharmaceutical Industry:

• Explanation of differences between statistical methods used for demonstrating comparability and when each is appropriate.
• A list of considerations for selecting an appropriate method.
• An approach for setting comparability criteria.
• A method to determine reasonable sample sizes.

Examples discussed in the course include: qualification of reference standards, comparison of manufacturing sites, partitioning of variation in gene and cell therapies, analytical bridging studies, tech and analytical transfer, scale-down model (SDM) qualification, and analytical similarity of biosimilars.

Methods discussed in the course include: side-by-side plots, statistical equivalence tests for means (TOST), non-inferiority of standard deviations, and quality ranges (min-max intervals, tolerance intervals, 3-sigma intervals, and risk-based side-by-side intervals).

Outline:

1. Comparability Studies
   • Comparability Basics and Definitions
   • Comparability Tools and Statistical Designs
   • CMC Applications
2. Comparability Tools
   • Considerations in Selecting an Approach
   • Statistical Tests
   • Quality Ranges
   • Comparison of Methods
3. Defining Acceptance Criteria and Power Calculations
   • Power and Statistical Power Calculations
   • Final Recommendations

Statistical Methods for Reliability Engineering and Survival Analyses:

This course covers the graphical and quantitative methods used reliability and survival analysis using JMP software. We will explore statistical methods to characterize and predict reliability and probability of survival. Topics include censoring, probability distributions—particularly Weibull and Lognormal, sample size determination, degradation/stability analysis, reliability with covariates/predictor variables, survival analysis, and fitting parametric survival models.

Statistical Methods for Reliability Engineering and Survival Analyses Objectives:

• Understand why and how to censor data from survival studies
• Be able to interpret nonparametric survival plots and models
• Determine the best probability distribution and parameter estimates to characterize reliability
• Be able to model the impact of predictor variables (covariates)
• Calculate sample size estimates for reliability demonstrations
• Model system performance with survival and parametric survival techniques

1. Introduction to Reliability Methods
• Censoring and event plots
• Functions and definitions: reliability, survival, hazard, cumulative
• Nonparametric survival methods
• Probability distributions for reliability analysis

2. Reliability and Survivability Models
• Survival analysis
• Reliability with a single predictor variable
• Reliability with multiple predictor variables using parametric survival models

3. Advanced Approaches
• Degradation and stability analysis
• Sample size determination
• Accelerated life tests