Statistical considerations on experimentation in animal behaviour studies

Abstract

One of the most crucial assumptions for a statistical analysis to be valid is that observations are independent, or that the dependence structure is known. Observations on behaviour obtained from different animals or at different times within a group of animals that are housed together will not have a known dependence structure. Therefore, such a group has to be considered as an observational unit and will be an experimental unit in experimentation. If statistical evidence on differences between treatments is required then it is necessary to have sufficient independently treated experimental units. The unit size, the construction of the units and blocks of units, the possibilities of change-over designs and of designs with different treatments within the same group of animals, the number of units and the amount of effort to be spent on each unit need to be thoroughly considered with respect to validity of statistical analysis and to optimal allocation of total research effort. Also to be considered is the possibility of using and planning meta-analyses and generalizing results to conditions in practice. It appears that the best strategy is using sufficient independent units for different treatments or treatment sequences and to make these units small or restrict effort per unit by sampling. Additional research on units of practical size may be needed for estimating the size of effects in practice. Heart-rate telemetry in deer and sheep R.M. Sibly, S. Price, G. Horalek and A.R. Jones Department of Pure and Applied Zoology, The University, Reading RG6 2/t,1, UK ABSTRACT Non-invasive methods of heart-rate telemetry will be reviewed and demonstrated, including skin electrodes, and ear and tail tags. The criteria used in evaluating system performance are: disturbance of behaviour; ease of fitting; reliability and durability; after-effects for the animal. Skin electrodes under a girth give best overall results. However, electrical noise resulting from electrode movement or muscle potential restrict usable information to about 85% of that collected. A common experimental design involves attaching and removing instrumentation from study animals at the start and end of each day. The effect of a treatment (e.g. loading into a vehicle) is studied by observing behaviour and heart rate for three periods: before, during and after the treatment. We have generally used 0.5 h for the before-treatment period and 2 h for after-treatment. In order to reduce the effects of social group and day the study of replicate groups, in each of which just one animal is instrumented, was considered to be an attractive proposition. However, considerations of practicality and time may restrict replication to within-group. It is necessary to know the effect of instrumentation itself on heart rate.Non-invasive methods of heart-rate telemetry will be reviewed and demonstrated, including skin electrodes, and ear and tail tags. The criteria used in evaluating system performance are: disturbance of behaviour; ease of fitting; reliability and durability; after-effects for the animal. Skin electrodes under a girth give best overall results. However, electrical noise resulting from electrode movement or muscle potential restrict usable information to about 85% of that collected. A common experimental design involves attaching and removing instrumentation from study animals at the start and end of each day. The effect of a treatment (e.g. loading into a vehicle) is studied by observing behaviour and heart rate for three periods: before, during and after the treatment. We have generally used 0.5 h for the before-treatment period and 2 h for after-treatment. In order to reduce the effects of social group and day the study of replicate groups, in each of which just one animal is instrumented, was considered to be an attractive proposition. However, considerations of practicality and time may restrict replication to within-group. It is necessary to know the effect of instrumentation itself on heart rate.