Autocorrelation of Timeseries

This Eco-Tool performs a variety of standard statistical explorations of a timeseries, generating an autocorrelogram, partial autocorrelogram, and power spectrum.

Input

Enter or browse to the location of a text file containing one or two columns, representing a timeseries. In a one-column file, the observations will be taken to come from a regularly-spaced sequence of times, which will be designated 1, 2, … etc. If there are gaps in the observations, the missing records should be included in the form of a non-numerical representation (such as the letter “n” — the exact letter does not matter). In a two-column file, the first column is of times, and the second column is of observations. Sequential observations should still be equally spaced in time, but missing observations do not have to be included.

Use my data:

Use 114-year record of Lynx canadensis population fluctuations from Mackenzie River (Elton and Nicholson 1942). To emulate analysis in Royama 1992, choose a maximum lag of 30, log-transformation, and the “efficient, biased” function below.)

Use generated random timeseries (white noise) of 100 steps.

Use generated one-dimensional random walk of 100 steps.

Extent of lags

Enter a value in the following box for the maximum time lag to be considered. It can be entered in one of two ways. If the number is in the interval 0 to 1, it will be interpreted as a fraction of the length of the data (thus, 0.5 would indicate a maximum lag of one half of the overall timeseries length). An integer value > 1 is interpreted as the number of time steps.

Analysis options

Log-transform data before analysis?

Form of autocorrelation function:

Why?)

References

  1. Elton, C. and M. Nicholson (1942) The ten-year cycle in numbers of the lynx in Canada. Journal of Animal Ecology 11: 215–244.
  2. Royama, T. (1992) Analytical population dynamics. Chapman & Hall, London.