ANOVA for temporal factor data - Data thinning to avoid autocorrelation/dependence?

KCG

New Member
#1
Hi all

I am conducting an analysis to assess the temporal occurrence of a species of whale in our survey area at different time scales using sightings per unit effort (SPUE, number of sightings per hour of effort) data. I am running separate ANOVA’s for each of the factors of Survey Week (18 weeks), Survey Fortnight (9), Month (4) and Year (4). I would like to compare the mean SPUE across each time scale to test if the mean SPUE changes in the area over the survey period and for what time intervals this might be significant. According to what I have read on the assumption of independence for ANOVA, the survey design is the main driver of whether the data is independent and data thinning can be used to ensure independence of samples i.e. removing re-sights of individually identified whales because successive observations of the same individual would be correlated in time. Thus far I have not included observations of re-sights of individuals at the Date level but my question is, do I need to remove re-sights of individuals within all time scales? If the same individuals are observed again in the same Survey Week, Survey Fortnight, Month or Year do they also need to be removed to ensure independence at those times scales? In one of my text books it states that serial autocorrelation will mostly affect individually referenced data, i.e. photographically identified individual whales in this case, and is caused either by physical limitations in movement (animals unable to transverse the study region within a single sampling interval). Given this information I feel that removing re-sights at the date scale is relevant but the animals can definitely transverse the study region (~500 km sq) at the Week, Fortnight, Month and Year time scales, so it sounds like I don't need to remove re-sights at those times scales. As a side note we don't have a huge number of re-sights within and between years so this would not result in a huge loss of data if I did remove those resights.

Thanks very much!