 |
||
Marie Ng, Post-Graduate FellowHong Kong, China
PhD, Quantitative Psychology; MS, Statistics
University of Southern California BS, Psychology Marie Ng puts IHME’s methods through the equivalent of a crash test. With training in both psychology and statistics, Marie questions the assumptions underlying health research and probes the weaknesses of analyses. Researchers might create a smooth curve in a trend line. For Marie, that can be a sign that the methodology is too disconnected from practical public health settings. In a practical setting, a country can go decades between health surveys, forcing researchers to fill in the gaps. Raw data can be coded incorrectly, requiring extensive revisions. Far from the smooth bell curve so often seen in research journals, trends can vary wildly over time. “We are trying to decide which method is best to generate the strongest results,” Marie said. “I am helping to identify the right method and then test that method as we go forward to make sure we are using it correctly.” While earning her Bachelor of Science in Psychology at the University of Washington in 2004, Marie found herself doubting some of the data underpinning major psychological studies. Those doubts have become a key part of Marie’s strategy for testing research methods at IHME. In the area of evaluations, Marie critically assesses statistical methods used to analyze time series cross-sectional data. In practical terms, that means she checks other people’s math. Marie has compared and contrasted statistical methods for IHME’s analyses of insecticide-treated bed net distributions in Africa. She also has analyzed the effectiveness of AIDS prevention programs in India. Estimating HIV prevalence trends is critical in monitoring the development of the epidemic and in evaluating the impact of various prevention programs, but solid estimates of HIV trends have been elusive. IHME researchers were struggling with the best method to calculate these estimates until Marie found a way to solve the problem. “IHME’s ultimate goal is to evaluate these interventions and find out which ones are working,” Marie said. “We also need to know if our evaluation methods are working, so we put them through some strenuous tests. If they don’t work, we find one that does.” Selected Publications:
|
