Impact of non-evenly sampled data on period estimates.

The reference data sets together with data sets from which 30% data points were randomly removed, were analysed with all the methods. The mean period and the absolute error are reported in the table. The reference sets consisted of 5 days of hourly-sampled data, to which walking noise was added at 80% level of the original amplitude. 1) The type of the base signal, (pul) pulse, (all) aggregated results from all the used shapes (cosine, pulse, double pulse, shoulder and moderate asymmetry). 2) Percentage of randomly removed points (time series with 30% points missing usually contained at least one 5-hour gap in the data). 3) Values of mean period (MP) and absolute error (AE)