Neurostatistics Working Groups

All PhD Students are encouraged to attend!

Wednesday, October 19, 2016.
Building 2, Room 426, 12:30 — 1:00 pm
Gabriel Torrealba, M.D.
The Costarican Stroke Registry Program

Cerebrovascular disease remains one of the leading causes of death and disability in Costa Rica. This is a retrospective, hospital-based registry, using systematic computer coding of data of all consecutive stroke patients admitted to the Hospital Calderon Guardia Acute Stroke Unit (HCG-ASU), since April 2009. In this study, we evaluated stroke patient profiles, risk factors and in-hospital complications and mortality. Preliminary results obtained from this registry will be discussed, as well as on-going and future projects related with this initiative.

Wednesday, October 26, 2016
Building 2, Room 426, 12:30 – 1:00 pm
Yorghos Tripodis, Ph.D.
Dynamic Factor Analysis for Multivariate Time Series: An application to cognitive trajectories

We provide statistical tools for analyzing Dynamic Factor Models (DFMs) using data that are typical in epidemiological studies with large number of participants and short non-stationary time series. Specifically, we develop an estimation algorithm, extending the classic EM algorithm, by developing an iterative two-cycle estimation process, following the steps of the ECME algorithm. This estimation method is flexible enough to be applicable in studies with multiple individuals, and short unequally spaced temporal information. We apply the DFM to a variety of neuro-psychological tests using data from the National Alzheimer’s Coordinating Center (NACC) study and estimate a smooth cognitive measure for each individual’s total cognition as well as measures for specific cognitive domains, such as memory, attention and language.  We show that by incorporating longitudinal information into the factor models we increase the accuracy of the estimates of change over time and consequently increase power to detect differences between groups.  The current model is applicable to data with short temporal component and unequally spaced observations. This is a particular strength of the estimation algorithm, since most of the observational studies on aging have these specific characteristics. The dynamic factor model is particularly useful when we are interested in finding differences in the rate of cognitive change between groups. This advantage can be used in future observational studies researching the heterogeneity in rates of progression of MCI and AD patients, or in future clinical trials that need to identify healthy participants at high risk of significant decline in cognition.