Analysis of presence/not detected data depends on the objective of the inventory effort. Suggested analysis methods for the given RIC objectives which will apply to most project-specific objectives are highlighted in Table 4. Each objective is discussed in further detail below.
Table 4. Survey objectives and analysis methods for presence/not detected data
|
RIC Objective |
Analysis methods |
Program |
|
Document species range |
|
|
|
Determine habitat associations |
|
|
|
Detect change in distribution over time |
|
|
1 See RIC Species Inventory Fundamentals manual for discussion of the negative binomial distribution.
Quantifying probability of detection of woodpeckers: The main purpose of these methods is to document species geographic ranges. From a statistical point of view it is important to attempt to quantify the detection probability (as a function of population density, population spatial distribution, detection probability, sampling effort, and other covariates) for each woodpecker species to allow a general estimate of the optimal amount of effort needed for surveys. Also, if an attempt is made to quantify probabilities of detection, a more statistically conclusive statement can be made about possible reasons for not detecting a woodpecker species as opposed to a simple "none were found" conclusion. A simple way to estimate probability of detection is through the use of the negative binomial distribution with data from relative abundance surveys. This procedure is detailed in the RIC Species Inventory Fundamentals manual, Section 5.
Documenting changes in woodpecker species distributions: If the objective is to detect changes in geographic distributions over time, we recommend a more intensive survey regime using relative abundance methods. This will allow a probability level to be associated with changes in distribution or apparent local extinction. A conclusion that species have become extinct in an area using presence/not detected methods will be difficult given that no estimate of survey precision is possible using current methods. More exactly, it will be difficult to determine that if a woodpecker species is not detected, whether it was due to lack of sample efficiency or actual demographic extinction.
Documenting habitat associations: If determining habitat associations is an objective, it will be important to document habitat types at the scale of woodpecker home ranges. This topic is addressed further in the RIC Species Inventory Fundamentals manual.
Detection rate may be used as an index of abundance, expressed either as the number of birds per visit (total of all call stations, total of all transects) or number of birds per call station (number of excavations per visit, number per transect). To account for observer bias, refer to suggestions made in the Controlling for observer bias section of this manual
If studies are designed appropriately the following general analysis methods can be used (Table 5).
Table 5. Survey objectives and analysis methods for relative abundance data
Objective |
Analysis method1 |
Programs2 |
|
Trends in abundance over time |
|
|
|
Comparison in abundance between areas |
|
|
|
Determine whether habitat modifications have altered population size |
|
|
1See the RIC Species Inventory Fundamentals manual for more details on analysis techniques
2See the RIC Species Inventory Fundamentals manual for more detail on software packages
As discussed below it may be possible to use program DISTANCE with encounter transect data to allow estimate of absolute density. Refer to Data Analysis of Absolute Density section in this manual and Section 5.4 - Absolute Abundance in the RIC Species Inventory Fundamentals manual for further discussion.
Difficulties with count data: One inherent problem with count data is that it is rarely normally distributed, which makes the application of parametric statistical methods risky, especially if sample sizes are low. Before data are used in parametric tests the assumption of normality should be tested. Transformations may make frequencies nearly normal in some cases. If data does not appear to be normally distributed then alternative data based methods using the negative binomial distribution (Program NEGTEST) exist. These newer methods also allow inference into population spatial distribution. A detailed discussion of analysis of count data is presented in Section 5.3 - Relative Abundance, of the RIC Species Inventory Fundamentals manual.
Trend analysis: The basic methodology for determination of population trends is linear regression. There are a variety of refinements to linear regression that can be used with data depending on sampling assumptions and other characteristics of the data. Refer to the RIC Species Inventory Fundamentals manual, Section 5.3 - Relative Abundance, for further discussion.
Comparison between areas: Parametric tests and other methods can possibly be used to compare areas if surveys are conducted concurrently. If surveys are conducted non-concurrently (such as in different years), then the results might be biased by population fluctuations. Refer to the RIC Species Inventory Fundamentals manual, Section 5.3 - Relative Abundance, and associated references for a thorough discussion of analysis of count data. Note particularly the potential for survey bias between habitat types and the potential problems that this may create in abundance comparisons. Different forest types and differences in vegetation density may affect both aural and visual detection of woodpeckers, creating perceived differences in abundance when there actually are no differences. Refer to Thompson et al. (1998) for further discussion.
Habitat based inference: Logistic regression or similar methods can be used to test for habitat associations, but this approach requires that habitat units be the primary sample unit as opposed to population units. Further modifications required for habitat inference are discussed in Sections 4.3 - Habitat Use and Selection and 4.4 -Making Habitat Inferences from Species Inventory of the RIC Species Inventory Fundamentals manual.