Categorising cheetah behaviour using tri-axial accelerometer data loggers: a comparison of model resolution and data logger performance

Table 5 Comparison of device performance (percent difference) for each behaviour in fine-, medium-, and coarse-scale models

Model	Behaviour	Percent difference (CEFAS-GCDC)	χ²	df	p	Better performing logger
Fine-scale behaviour model	Crouch	− 13.7	5.21	1	0.022	GCDC
	Lie	− 2.7	52.85	1	< 0.001	GCDC
	Sit	− 6.9	10.37	1	0.001	GCDC
	Stand	+ 2.6	11.63	1	< 0.001	CEFAS
	Head movement	− 10.8	8.51	1	0.004	GCDC
	Crouching stalk	+ 13.1	0.51	1	0.474	n.d
	Lying stalk	+ 0.7	1.14	1	0.286	n.d
	Sitting stalk^a	− 3.7	0.00		0.380	n.d
	Standing stalk	− 1.4	0.13	1	0.722	n.d
	Walking stalk	− 6.9	7.52	1	0.006	GCDC
	Trotting stalk	− 8.9	4.92	1	0.027	GCDC
	Walk	− 2.1	2.00	1	0.158	n.d
	Trot	− 16.1	15.40	1	< 0.001	GCDC
	Canter	− 3.8	1.91	1	0.167	n.d
	Gallop	+ 8.8	19.46	1	< 0.001	CEFAS
	Pounce	− 51.2	11.40	1	< 0.001	GCDC
	Mean difference	− 6.4	20.62	1	< 0.001	GCDC
Medium-scale behaviour model	Head movement	− 3.7	0.84	1	0.360	n.d
	Crouching stalk	− 2.6	< 0.001	1	1.000	n.d
	Lying stalk	+ 0.9	1.57	1	0.211	n.d
	Sitting stalk^a	− 7.4	0.00		0.141	n.d
	Standing stalk	− 0.9	0.03	1	0.872	n.d
	Walk	− 1.5	0.89	1	0.344	n.d
	Trot	− 15.7	14.88	1	< 0.001	GCDC
	Canter	− 4.3	2.52	1	0.112	n.d
	Gallop	+ 9.6	23.15	1	< 0.001	CEFAS
	Sedentary	− 0.4	2.25	1	0.134	n.d
	Moving stalk	− 10.0	23.95	1	< 0.001	GCDC
	Mean difference	− 3.3	5.52	1	0.019	GCDC
Coarse-scale behaviour model	Head movement	− 11.0	8.40	1	0.004	GCDC
	Active	− 1.4	3.11	1	0.078	n.d
	Inactive	+ 0.5	4.42	1	0.035	CEFAS
	Mean difference	− 4.0	0.87	1	0.350	n.d

Negative “percent difference” values indicate that GCDC devices performed best for categorising a given behaviour, whereas positive values indicate that the CEFAS logger was better
“n.d.”, no significant difference
^aFisher’s Exact test statistics (Odds ratio in lieu of χ²)

ISSN: 2051-3933