Reduced-representation sequencing methods have wide utility in conservation genetics of non-model species. Several methods are now available that reduce genome complexity to examine a wide range of markers in a large number of individuals. We produced two datasets collected using different laboratory techniques, comprising a common set of samples from the greater bilby (Macrotis lagotis). We examined the impact of differing data filtering thresholds on downstream population inferences. We found that choice of restriction enzyme and data filtering thresholds, especially the rate of allowable missing data, impacted our ability to detect population structure. Estimates of FST were robust to alterations in laboratory and bioinformatic protocols while principal coordinates and STRUCTURE analyses showed variation according to the number of loci and percent missing data. We advise researchers using reduced-representation sequencing in conservation projects to examine a range of data thresholds, and follow these through to downstream population inferences. Multiple measures of population differentiation should be used in order to fully understand how data filtering thresholds influence the final dataset, paying particular attention to the impact of allowable missing data. Our results indicate that failure to follow these checks could impact conclusions drawn, and conservation management decisions made.
