The source of ENSO data from NIST

Question

I'm using ENSO dataset from the NIST Statistical Reference Datasets as a test for nonlinear regression code. The data are monthly averaged atmospheric pressure differences between Easter Island and Darwin, called Southern Oscillation Index (SOI). In the test, the data are fitted with three sine cycles (like this).

I got curious about the Southern Oscillations and I started looking for more SOI data. I learned that there are a few different methods of how to calculate the SOI, but none of the other sources that I found (e.g. here) gives the same numbers are in the NIST StRD.

There is a reference in the dataset, Numerical Methods and Software by Kahaner, Moler and Nash, where I found only that the data are for years 1962-1975. So my question is not really about statistics: how the ENSO dataset is related to the SOI data from other sources.

Answer 1

You would not expect your two sets of numbers to be the same: the NIST numbers seem to be have been published to test statistical software while the Australian BOM numbers are designed to identify whether the ENSO cycle is above or below average.

In particular the BOM numbers have been standardised to an index to have monthly seasonal factors removed, and for the remaining numbers to have a mean of around 0 and standard deviation around 10. They may also have had corrections or improved sources used since the NIST numbers were published.

The NIST have had not had these proceses applied: in fact one of the purposes of the illustrated analysis is identifying the mean (coefficient b1) and the magnitude of the annual cycle (coefficients b2 and b3). It is possible that the NIST numbers are in real physical units (perhaps of pressure).

You might have hoped that there was a linear relationship between the two sets of numbers, for example between the January BOM numbers and data points 1, 13, 25, ... , 157 of the NIST numbers. I did not spot it, though I did not try very hard.