I regularly use time series of total large mammal and waterbird counts, to indicate population trends (see for example Scholte 2006, 2013 both in my RG profile). I often add trend-lines (linear or other regression, through a simple Excel exercise) to illustrate them. However, I receive comments that presenting such time series without ‘statistics’ (not further specified), lacks (in my own words) scientific rigor. We are currently working on modeling antelope population trends inspired by Joseph Ogutu’s work in E.Africa (‘using a flexible multivariate semi parametric generalized linear mixed model with a negative binomial error distribution and a log link function. The model smoothes the population trend, specified a cubic B-spline, etc.’ Ogutu et al. 2011), but this only works with sufficient observations and is rather cumbersome.
My question is two-fold:
- Am I right that for total counts (contrary to transect or other partial counts for which limits of confidence can be calculated) ‘statistics’ makes little sense?
- What are alternative ways to present total count time-series allowing an appreciation of their reliability and/or trend ?
Article Waterbird recovery in Waza-Logone (Cameroon), resulting from...
Article Population trends of antelopes in Waza National Park (Camero...