THOUGHT LEADERSHIP

This is the last in a series of three posts on missing data imputation.  In the first post, we reviewed the underlying assumptions and limitations of single imputation methods, in particular, last observation carried forward (LOCF).  In the second post, we considered more sophisticated data imputation methods for longitudinal data, Mixed Models for Repeated Measures (MMRM) and Multiple Imputation (MI).  An important assumption for both MMRM and MI is that the data are ‘missing at random’ (MAR), if not ‘missing completely at random’ (MCAR).  But what if this assumption is not true for all the missing data?

Last October, Susan Boquist, Associate Director, Statistical Programming, traveled to CDISC International Interchange in San Diego, California to present the poster, “Creating Harmonious SDTM Domains”.  The poster was created in collaboration with Principal Statistical Programmers: Elena Prosekova, Natalia Quinn and Sofia Tyryshkina. 

Mapping raw data into standardized, tabulated data sets can be a daunting task to undertake, especially if one has not done it before.  Even seasoned mappers could use the help of some additional ideas and tips.  The poster shared the authors’ thoughts, suggestions and examples to help turn a cacophony of raw data into a standardized symphony that will flow into analysis.

Due to the worldwide pandemic, the CDISC European Interchange, originally scheduled to take place in person in Berlin, Germany, was presented completely online. Susan Boquist, Associate Director, Statistical Programming, took advantage of the accessibility and logged in at 3:00 am her time both mornings. Though it may have become an April Fool’s Day to remember, with only a few weeks’ notice, the organizers were able to make the format transformation flawlessly. The conference schedule was consolidated into one track. A networking app was employed to allow for attendees to easily communicate with each other, ask questions of the presenters, and view and vote on the poster presentations. Training and workshops were rescheduled for another time. It was so successful that everyone hopes the organizers will consider adding a virtual option for future conferences. 

In a previous post on missing data imputation, we reviewed the underlying assumptions and limitations of last observation carried forward (LOCF). This method was widely used in the past because of its straightforward application, ease of understanding, and often incorrect assumption that it is a conservative approach. Over the years, more sophisticated data imputation methods for longitudinal data have been developed that have advantages over the single imputation methods, such as LOCF, baseline observation carried forward (BOCF) or worst observation carried forward (WOCF).  We will briefly review two of these methods, mixed models for repeated measures (MMRM) and multiple imputation (MI).