R-Package: FFD - Freedom From Disease using Two-Stage Sampling

Introduction

In practice, when conducting surveys to substantiate freedom from disease in large populations two-stage sampling strategies are often used in order to account for herd-level clustering of diseases. With two-stage sampling the questions that need to be addressed are a) how many herds need to be sampled and b) how many animals per herd need to be tested. Using a modified hypergeometric formula the optimal sample sizes can elegantly be computed, while incorporating imperfect diagnostic tests and finite populations; see Cameron and Baldock (1998a,b).

In the package FFD, tools for calculating optimal sample sizes (on animal and herd level) using sampling strategies "individual sampling" or "limited sampling" (see Ziller, et al., 2002) are implemented.

• Individual sampling: The idea behind individual sampling is that the number of animals to test is computed for each herd size in order to guarantee a constant herd sensitivity independent of the herd size, i.e., for each herd the probability of finding a possibly present disease is the same.
• Limited sampling: With limited sampling a pre-fixed number of animals is tested for each herd (=sample limit), irrespective of the herd size. This results in a herd sensitivity that varies with the herd size, i.e., the smaller the herd the higher the probability of finding a possibly present disease.

Both these sampling strategies have their merits, which can be analyzed and compared using the package FFD. Cost optimal sampling strategies can be determined finding the optimal herd sensitivity or sample limit, respectively. The package furthermore includes tools for evaluating the a-posteriori confidence corresponding to a specific sample of herds according to Kopacka et al., 2013. The package provides an easy-to-use graphical user interface (GUI). More advanced users can, however, make use of the provided S4-classes and methods, as well as the underlying conventional functions in order to fully make use of the maximum flexibility offered.

Risk-based sampling

The latest release of the package FFD now offers the possibility to use risk-based sampling. Targeting high-risk population strata, the overall sample size and the cost of the surveys can be reduced, while still maintaining the same overall significance level.

Features Easy-to-use graphical user interface S4-classes and methods, as well as conventional functions for maximal flexibility Calculation of sample sizes for strategies "individual sampling" and "limited sampling" Diagnostic tools for the optimization of efficiency and costs A-posteriori calculation of the alpha-error corresponding to a specific sample Sampling methods using either a fixed number of herds or dynamically updating the achieved significance of the sample and randomly sampling herds until a specified significance level is met Risk-based sampling

Package Installation

To install this package directly within R type:

install.packages("FFD", repos="http://R-Forge.R-project.org")

or download source code or binary from R-Forge.

Download and Links

• Download source code/binary from R-Forge
• Package vignette [pdf]
• FFD on CRAN

References

• Cameron A.R., Baldock F.C. (1998a). "A new probability formula for surveys to substantiate freedom from disease", Preventive Veterinary Medicine, 34, 1-17.
• Cameron A.R., Baldock F.C. (1998b). "Two-stage sampling surveys to substantiate freedom from disease", Preventive Veterinary Medicine, 34, 19-30.
• Kopacka I., Hofrichter J., Fuchs K. (2013). "Exact alpha-error determination for two-stage sampling strategies to substantiate freedom from disease", Preventive Veterinary Medicine, 109, 205-212.
• Ziller M., Selhorst T., Teuffert J., Kramer M., Schlüter H. (2002). "Analysis of sampling strategies to substantiate freedom from disease in large areas", Preventive Veterinary Medicine, 52, 333-343.