A new open source tool called the Anchored Multiplier can help researchers more accurately estimate the size of a population when they have at least two estimates calculated using different methods.
Developed by Paul Wesson, a post-doctoral fellow at UCSF, IGHS faculty Willi McFarland and Ali Mirzazadeh and software engineer Cong Charlie Qin, the tool enables researchers to incorporate prior information along with multiple estimates from empirical data in the calculation. It uses a Bayesian framework to weight individual estimates according to their precision (i.e. the width of their confidence intervals), creating a single “best” estimate from multiple, possibly disparate, estimates.
Typically, researchers calculate the median of multiple estimates to get a final estimate. But when Wesson conducted a systematic review of the population-size-estimation literature looking at studies that implemented at least two size estimation methods, he found that most produced discrepant results.
“I was not satisfied with this approach, as it treats all estimates as equally valid,” Wesson said. “Some estimates may warrant contributing less to the final estimate because they are less precise or perhaps biased.”
Wesson and his colleagues developed the Anchored Multiplier for investigators who may not have a strong background in statistics or Bayesian analysis. People working in Ministries of Health or in public health departments may also find it useful.
“The only way to get a truly accurate estimate of population size is through a complete census,” Wesson said. “In the absence of this gold standard, however, the next best approach is to estimate the population size using multiple methods; thereby assessing the consistency of estimates or the range of potential estimates,” he said. “The Anchored Multiplier allows researchers to synthesize all available data into a single ‘best’ estimate.”