Displaying Ordinal Data (Bidirectional)

The following stacked bar graph has many advantage:

• The most intense ratings are stacked nearest the axis, with decreasing intensity away from the axis.  Hence the top of any bar corresponds to the % of subjects with at least that intense of a response.  
• E.G.: at a glance: For the EOS group A observation, about 10% scored "Very Much Improved" (top of the dark green), while 30% scored at least Much Improvement (top of the light green), and 30% any improvement (NB: no one scored "Minimally Improved")
• The above axis item correspond to improvement, and below axis items correspond to worsening
• E.G.: In the Day 28 Group A observation, few scored Very Much Improvement or Worsening, and about equal numbers scored improvement as worsening. 
• Ratings of "No Change" are implied, but not shown.
• E.G.:  For Day 28 Group A, about 30% improved, 40% worsened, leaving 30% with no change.

Generic Description (for Statistical Analysis Plan)
For birectional ordinal data (such as CGI-I), the same concept as for unidirectional ordinal data is used but with desirable scores (e.g., improvement) plotted above the horizontal axis, and undesirable scores (e.g. worsening) plotted below the horizontal axis. For either category, the most extreme ratings will be closest to the horizontal axis, and assigned the most intense colors. This will allow an immediate visual impression of the relative proportions of patients who improved versus those who worsened, and by how much.

Project Design Mapping: an Organizational Methodology

I've been summarizing the best approaches I've found to organize the many design elements of a clinical project & packaging it into a slide set.

Why? For example, when designing a set of clinical studies, we might base an endpoint on similar studies in the literature, have different variations of the endpoint in different studies, use various analytic methods, etc, all of which is documented in a long series of text documents (see list on the slide below). Likewise for objectives, visits, assessments, inclusion/exclusion criteria, tables, figures, and listings, etc.

Over years, I've developed an approach to organize each design element (and variation thereof) among a set of studies into a spreadsheet. Essentially, the spreadsheet is a map to all of the design elements in a clinical program: instead of having to search through multiple documents for details, it is all in a single spreadsheet. For instance, if the definition of "Metabolic Syndrome" changes in the middle of a program, the design map shows at a glance which studies used which definition.

(NB: My daughter made these drawings)

Variance vs Standard Deviation: which to use?

Question: "It is difficult for me to comprehend the difference between standard deviation and variance after having good amount of reading. Can you pl help me." (from LinkedIn)
My answer:

Variability
Show Standard Deviation
Calculate Variance

Ramblings about clinical statistics

When I started doing statistics for psychiatric clinical trials, I visualized my role as being on the top of a pyramid: standing on all the information & data from the study, I was privileged to 1st see & understand the results of the study.

Later, I visualized my role as being at the bottom of a funnel: each of thousands of numbers coming out of the funnel was the distillation of a tragic story entering the funnel, a summarization of decades of suffering for a patient and their family.

(I recently wrote this on Facebook)

Simulating randomized withdrawal studies for prediction & management

Title of talk: Signs of the Timings: Predicting Time of Completion in Multiphase Survival Trials

Paper as published in the conference proceedings: Predicting Time of Completion in Multiphase Survival Trials

Abstract:

Studying maintenance of clinical effect typically requires clinical response for a minimum amount of time on treatment before randomization. If randomized, patients are then followed until treatment failure or withdrawal, and the trial halted after a pre-specified number of events. For ethical and cost reasons it is desirable to minimize the number of patients enrolled and randomized, and to predict the time of the last event under multiple scenarios.

We describe a data-driven stochastic simulation for two such trials in which: Each phase is modeled as a competing event process; Distributions of event times are derived from Kaplan-Meier survival curves from available data; Parameter uncertainty is modeled based on K-M survival estimates; Withdrawals and events occur at similar overall rates, though at different times; Predictions are updated as information is accrued.



Presented to:
Joint Statistical Meeting of the American Statistical Association
Seattle, WA
August, 2006

Delaware Chapter of the American Statistical Association
Dinner Meeting: September 21, 2006

Speaker: Dennis E. Sweitzer
AstraZeneca