Analysing variance

#1
Can anyone direct me to a suitable statistical test for the following?

I want to evaluate the mean and variance for data from a nested survey design.

The data are counts of habitat categories along a transect. Each transect is divided into 4 sub transects to provide statistical replication. (Counts can be evaluated as percentage covers for each transect- but this may not be relevant to this test)

This project aims to validate the survey method by looking at precision and variability. Nested experimental design to achieve this is as follows:

2 sites
both visited by two teams of 6 surveyors each
3 transects deployed on each site
each transect is surveyed by each of the 6 surveyors in each team

The aim is to examine inter-surveyor variability as they each survey the same transect.

The attached excel sheet provides an example of the type of data.

What is the most appropriate model to use here. I have been looking at ANOVA and GLM, but as the data is not of the predictor/ response type, is this appropriate. Also there are some missing values in the data, how can these be dealt with?

Grateful for any pointers in the right direction, as I'm lost in a maze of statistics at the moment.

Thanks
 

rtd

New Member
#2
This project aims to validate the survey method by looking at precision and variability.
I do not know how to validate the precision, when there is no exogenous "right" answer.

You can compare the variability by comparing the variances of the surveyors answers. If one is significantly lower you could call it a superior method... (perhaps only if there are no other reasons why it could be lower, e.g. more motivated surveyors?) If you do a web search on "differences in variance" you will find many tutorials explaining how to compare two variances.

HTH
Michael