Statistical Power Analysis in ODA, CTA and Novometrics (Invited)

Nathaniel J. Rhodes

Chicago College of Pharmacy, and the Pharmacometrics Center of Excellence, Midwestern University

Statistical power analysis simulation results are provided for determining the “worst-case” sample size assuming minimal measurement precision and relatively weak or moderate effect strengths. In simulated trials with relatively weak effects (ESS = 24%), greater than 80% and greater than 90% power is achieved for equal sample sizes for unit-weighted analyses of one comparison at n = 70 and n= 90 subjects per group. For simulated trials with moderate effects (ESS=48%) greater than 80 and greater than 90% power was achieved for the same conditions at n = 20 and n = 25 subjects per group. A simulation-based approach is useful for determining the “worst-case” sample sizes in various applications.

View journal article

Implementing ODA from Within Stata: An Application to Estimating Treatment Effects using Observational Data (Invited)

Ariel Linden

Linden Consulting Group, LLC

In this paper, I demonstrate how treatment effects in observational data can be estimated for both binary and multivalued treatments using the new Stata package for implementing ODA. Matching and weighting techniques are implemented and ODA results are compared to those using conventional regression approaches.

View journal article

 

Implementing ODA From Within Stata: An Application to Data From a Randomized Controlled Trial (Invited)

Ariel Linden

Linden Consulting Group, LLC

In this paper, the new Stata package for implementing ODA is introduced by reanalyzing data from a study by Linden and Butterworth (2014) that investigated the effect of a comprehensive hospital-based intervention in reducing readmissions for chronically ill patients. In the original analysis, negative binomial regression was used to evaluate readmission rates and emergency department visit rates at 30 and 90 days, and no treatment effects were found. However, ODA is a superior analytic approach because of its insensitivity to skewed data, model-free permutation tests to derive P values, identification of the threshold value which best discriminates intervention and control groups, use of a chance- and complexity-corrected indexes of classification accuracy, and cross-validation to assess generalizability of the findings.

View journal article

 

Software Sale Live

Softwares MegaODA and CTA, as well as a .pdf  of the book Maximizing Predictive Accuracy, are for sale on the Purchase page for $9.99.


 

To run MegaODA and CTA softwares within Stata refer to these pages. Dr. Ariel Linden created and published the Stata programs for these softwares.

ODA, LLC will publish instructional videos to YouTube on these softwares regularly.