INQUA Working Group on Data-Handling Methods

Newsletter 2: June 1989

Polsta Interactive Time Series Analysis

POLSTA is a computer programme designed to simplify the preparation, plotting, and analysis of time-based observations. Short for "Pollen Time Series Analysis", POLSTA is applicable to a wide range of data, including pollen counts, tree rings, meteorological readings, and economic records. It was designed by Dr. David G. Green of the Australian National University to meet three common needs of scientists today.

1. To cope with large data sets, especially those consisting of many separate series of observations; the standard POLSTA package can handle up to 200 series of 500 observations; (versions that accommodate larger data sets can be supplied).
2. To reduce the time needed to perform routine tasks with your data. For instance, using POLSTA it is possible to generate a plotted pollen diagram from a raw counts in a matter of minutes.
3. To make it possible for non-statisticians to use powerful time series statistics and modelling techniques.

Major Subroutines

• Data Handling Up to 200 taxa and 500 samples. Any input format acceptable. Up to five ancillary series, not to be analyzed, can also be accepted. Data set can be output in any format to disk file thereby eliminating the time needed to set up data appropriately in later sessions. Data series are stored in a direct-access file; simple commands are used to define a workspace containing copies of up to 20 data series at one time. Program can create new data series.
• Algebra Create or modify data series in response to simple formulae. Most standard arithmetic operations are possible plus integration, moving or changing segments of data, and generating models such as polynominials, sine waves and random series.
• Chronologies Sample ages, and associated sediment deposition rates, can be estimated either by finding the power curve of best fit to known dates, or by fitting straight line segments between them. More complex chronologies can also be constructed by using the algebra routine.
• Data Plotting A variety of data plotting is available. Conventional diagrams use depth on one axis and data series (one by one) on the other axis. Scattergrams may be plotted for pairs of series. Data may be plotted as points, lines, bars, or histograms. Rough plots of series can also be printed at the terminal or line printer.
• Detrending and Filtering Performed algebraically or by best fits to portions of a series of the appropriate linear, power or logarithmic curve. Special commands are available for filtering series by moving average, autoregression and Nth differences.
• Elementary Statistics Means, standard deviations, and extrema can be computed for any segment of the data.
• Time and Frequency Domain Analysis Serial and cross-correlegrams, power spectra, and coherency/phase spectra. Select the spectral window and check aliases, bandwidths, and confidence intervals of cycling frequencies detected.
• Multiple Regression Linear combinations or values in up to 20 independent series. Non-linear and lagged relationships are found by generating the appropriate independent series algebraically. Observed and estimated values of the dependent series can be plotted and the residuals computed.
• Sequence Splitting Search for significant changes in mean and variance - an alternative to zonation in the interpretation of stratigraphic records. POLSTA displays the location of sequence splits and gives statistics for the subsections identified.
• Special Purpose Features Some optional routines may be available upon request. Please explain your needs in detail.

Information available from: Anutech Pty Ltd., Childers Street ACTON ACT 2601, GPO Box 4, Canberra ACT 2601. Telex AA 62760 Natuni.