Evaluation of analystical procedures for density estimation from line-transect data: data grouping, data truncation and the unit of analysis

Abstract

We examined three aspects of line-transect analytical procedures: data grouping, data truncation and the use of individuals or clusters as the analytical unit. Bias and precision of density estimation in relation to various levels of these factors were assessed for 4 types of line-transect estimator (simple parametric, generalised parametric, non-parametric and quasi-strip) using line-transect survey data from macropod populations of known density. The effect of data grouping on bias and precision varied between estimators. Bias was stable across all grouping levels tested for the simple parametric estimator, and stable across aU but the coarsest grouping level for the generalised parametric and non-parametric estimators, but varied substantially across the range of levels tested for the quasi-strip estimator. Precision improved as the number of grouping levels increased for all estimators tested, but the extent of improvement varied between estimators, and for the estimator most affected, improvement was marginal beyond intermediate grouping levels. Density estimates were generally more accurate and precise when analysed in ungrouped form than in grouped form. No effects of truncation on bias or precision were detected. Varying the analytical unit did not affect bias, but precision was significantly lower for cluster analysis than individual analysis for all estimators.