Assessing the utility and limitations of accelerometers and machine learning approaches in classifying behaviour during lactation in a phocid seal

Table 5 Variance and repeatability estimates for individual ID

Behaviour	Variance	SE	Feat. var.	adj.-R
Rest	0.042	0.019	stX	0.373
Alert	0.042	0.020	stZ	0.295
Presenting/Nursing	0.122	0.053	stX	0.600
Locomotion	0.037	0.018	stZ	0.248
Comfort Movement	0.212	0.096	Pitch	0.619
Flippering pup	0.058	0.029	Roll	0.925*

Variance explained by individual ID across pooled data of top feature variable, stZ. Variance and standard error (SE) of the random effect of individual ID were estimated over 1000 bootstrapped samples within a generalized linear mixed effects model using the ‘rptR’ package in R [63], adjusted for maternal post-partum mass for repeat capture females (n = 11). No variance was explained by the random effect of year. Repeatability (adj.-R) for each behaviour was estimated with specific feature variables (Feat. Var.) corresponding to each behaviour using the same mixed effect model structure. Those repeatability estimates that were significant as determined through likelihood ratio tests are labelled with an (*)

ISSN: 2050-3385