I prepare a census with the aim to estimate the densities of bonobos, elephants and ungulates. Our census protocol combines the recce and transects. Following the same protocol we obtained densities of Bonobo in 2012. But the coefficient of variation was so high. Can anyone help me to know what to do to reduce the coefficient of variation?
Probably if you try bigger samples with much more transects as sampling units and also standardize times, locations and samplers
Probably the best thing to do now is to discuss comprehensively the results assigning them to your sampling design
Keep in mind that once you have a decent enough sample size to get reliable estimates of mean and standard deviation, the CV will not change. It is the standard deviation divided by the mean, and neither of those changes with sample size.
One idea however is it sounds like you could rethink your sampling design. Perhaps changing the scale of your sampling units, or doing some stratified sampling. These can bring down your variance.
Dear Alain, it is not clear what you mean with too high. What statistical test did you do or want to do?
Thank you Christian W W Pirk and all for your answers.
I conducted the analysis for bonobo densities using Distance software. By using several different models of detection function for each density calculating (Half-normal cosine, cosine Uniform, Uniform single polynomial, Half-normal Hermit polynomial, Hazard rate cosine), I found a CV ranging between 34 and 40% this is too High. So it is a possibility to reduce the CVs I'm looking for the next study.
- Badges
- Science topic
- Similar topics
- Agricultural Science
- Agronomy
- Crop Science
- Crop