Autoassociation is analogous to autocorrelation, but for a sequence of binary or nominal data coded as integer numbers.
For each lag, the autoassociation value is simply the ratio of matching positions to total number of positions compared. The expected autoassociation value for a random sequence is calculated according to Davis (1986). The resulting p values (two-tailed) can be shown as a function of lag.
The multiple testing issue arises for the set of p values.
The test above is not strictly valid for “transition” sequences where repetitions are not allowed. In this case, select the “No repetitions” option. The p values will then be computed by an exact test, where all possible permutations without repeats are computed and the autoassociation compared with the original values (one-tailed). This test will take a long time to run for n>30, and the option is not available for n>40.
For computational details, see the Past manual.
Missing data supported.
Reference
Davis, J.C. 1986. Statistics and Data Analysis in Geology. John Wiley & Sons.