Measurements. |
What are you measuring?
Ethologists often collect data that fall into mutually exclusive categories. The reliability training process that we undertake varies depending on the type of data collection.
Tasks and techniques
Ethologists often conduct specific tasks or techniques to generate data. Although some of these may seem straightforward on the surface, there may be variation in execution. Training and evaluating reliability in performance of tasks and techniques is recommended. Examples include:
- Taking photographs
- Weighing animals
- Measuring feed intake
- Using a precise instrument, e.g. an algometer to quantify tissue sensitivity in pain research
Categorical and continuous data
The data that we collect are ultimately categorical or continuous.
-
Categorical data
-
Continuous data
<
>
Categorical data are data collected in discrete bins, like yes/no, or on a discrete scale, e.g. 1, 2, or 3. Discrete outcomes can only be specific, set values. Categorical variables are either nominal or ordinal.
Nominal outcomes are not ordered (e.g. hair color - one color isn't considered greater or less than another color). If there are only 2 categories (yes/no), then they are considered binary.
Ordinal outcomes follow an order or rank (e.g. score 1 < score 2 < score 3). While positions in ordinal data matter, the differences between positions may not be meaningful (e.g. score 2 isn't necessarily twice as "good" as score 1).
Examples include:
If you plan to collect categorical data, you also need to develop robust definitions for your data before beginning your experiment. This will influence how you approach training.
Nominal outcomes are not ordered (e.g. hair color - one color isn't considered greater or less than another color). If there are only 2 categories (yes/no), then they are considered binary.
Ordinal outcomes follow an order or rank (e.g. score 1 < score 2 < score 3). While positions in ordinal data matter, the differences between positions may not be meaningful (e.g. score 2 isn't necessarily twice as "good" as score 1).
Examples include:
- Presence/absence of a behavior (e.g. panting)
- Hygiene scores (clean, somewhat dirty, filthy)
- Wound scores
If you plan to collect categorical data, you also need to develop robust definitions for your data before beginning your experiment. This will influence how you approach training.
Continuous data are collected on an interval or ratio scale that can take on an unlimited number of values between 2 points. The difference between points on a scale is meaningful. For example, 35 degrees Celsius is warmer than 17 degrees Celsius, and an animal that is eating for 20 minutes is engaging in that behavior for twice as long as one that eats for 10 minutes.
Examples include:
If you plan to collect continuous data, you will decide on an appropriate sampling schemes (e.g. continuous vs. time sampling; Bateson and Martin, 2021) for your variables of interest, and develop robust definitions for your variables. These choices will influence how you approach training.
Examples include:
- Duration of a behavior (e.g. time spent eating, proportion of observations spent grooming)
- Latency to approach an object
- Frequencies or counts of an outcome (e.g. number of vocalizations)
If you plan to collect continuous data, you will decide on an appropriate sampling schemes (e.g. continuous vs. time sampling; Bateson and Martin, 2021) for your variables of interest, and develop robust definitions for your variables. These choices will influence how you approach training.
