. 2014 May 8;9(5):e96462. doi: 10.1371/journal.pone.0096462

Table 5. Impact of sampling frequency on period estimates (data sets with walking noise).

Shape¹	Method	0.1 (3.0)²	1.0 (3.0)²	0.1 (5.0)²	1.0 (5.0)²	2.0 (5.0)²	2.0 (10)²
cos	EPR	23.62	23.63	23.93	23.89	23.88	23.97
cos	MFF	23.96	23.88	24.02	23.99	23.97	24.00
cos	NLLS	23.89	23.84	23.98	23.96	23.96	24.00
cos	MESA	23.81	23.87	23.95	23.94	23.95	23.99
cos	LSPR	23.91	23.87	24.01	23.99	23.98	24.00
cos	SR	24.2	24.25	24.11	24.09	24.07	24.01
asym	EPR	24.54	24.48	24.06	24.1	24.09	24.08
asym	MFF	24.16	24.05	24.09	24.06	24.03	24.09
asym	NLLS	24.11	24.11	24.11	24.12	24.12	24.09
asym	MESA	24.00	24.00	24.01	24.02	24.03	24.06
asym	LSPR	24.65	24.55	24.28	24.28	24.24	24.13
asym	SR	24.53	24.57	24.23	24.22	24.21	24.07
all	EPR	24.02	23.99	24	23.98	23.98	24.01
all	MFF	24.02	23.96	24.03	24.01	23.99	24.02
all	NLLS	23.98	23.92	24.01	23.99	23.99	24.01
all	MESA	23.89	23.91	23.96	23.95	23.96	23.99
all	LSPR	24.22	24.14	24.1	24.09	24.07	24.03
all	SR	24.40	24.40	24.16	24.14	24.13	24.01

Data sets with different time intervals and selected durations were analysed using all the methods and the mean period value is reported in the table (standard deviations are omitted for clarity). Data sets were created by adding walking noise of 80% of the original signal amplitude to the templates of different duration and time interval between points. The underlying period was 24.08 h for asym data and 24.00 h for the other signals. 1) The base shape of the signal: cosine (cos), pulse (pul); double pulse (dpl); shoulder (shl) and moderate asymmetry (asym), (all) represents aggregated results from all the sets. 2) the time interval (sampling frequency) in the data set and in brackets the data duration. See SI Table S5 for the full table.