How to interpret data
Choosing how to edit, process and interpret data.
-
Protocols used.
Clearly document the steps you undertake to edit, process and interpret the data. Develop standardised processes for dealing with data. For example, how did you deal with missing data, imprecise bird counts or unidentified bird species?
-
Repeated measures.
If your dataset includes multiple measures within a sampling unit, then team up with a statistician to make sure these data are processed correctly.
-
Sampling effort.
If sampling effort (e.g. number or duration of point counts, or length of transects walked) varies among your sampling units, you need to account for this in your analysis.
-
Imperfect detections.
Consider whether, at the analysis stage, you need to correct for any birds or species that were present within the sampling unit.
-
Contextual information.
Consider whether it is necessary to account for any variation in the environmental conditions in which the counts were carried out (e.g. differences in the habitat types within your study units, changes in observers or weather conditions) to help interpret the patterns observed.
-
Uncertainty measures.
Provide an estimate of uncertainty or error for the measures you are reporting to help people interpret the results. Basic summary statistics can be used to provide this information.
-
Avoid false alarms.
Carefully interpret observed patterns in the context of the ecology of the species or bird community being monitored. False alarms can result from statistical noise and fluctuations in trend data, particularly when using ‘alert’-based systems.
-
Invite a review.
Having a fresh set of eyes, preferably an external expert, to review your analysis can help you verify that your analytical approach is correct and that you haven’t misinterpreted the results.