sine and cosine functions (see Gottman, 1981). This implies that investigators can examine the relationship between two time-series irrespective of whether or not cyclicity exists within the data.
Like the sections on lag-sequential analysis, and log-linear models, this section on time-series analysis is intended only to introduce the concepts of this method of data analysis, and to illustrate how this approach is applicable to family interaction data. For greater depth, interested readers are directed to Chatfield ( 1984), Gottman ( 1981), and Kendall ( 1973). Gottman's ( 1981) text is directly related to the social interaction perspective, and the other books provide the breath needed to fully understand the analysis of time-series data.
To assess the relevant temporal structure in sequential data requires an initial decision of whether to pool subjects, or to analyze subject-by-subject. If the latter option is taken, and there are sufficient data to warrant confidence in the stability of the statistics, the investigator can use either lag-sequential analysis or time-series analysis. If the investigator elects to pool the data, then log-linear models offer an array of procedures for examining the data's temporal structure. Regardless of the option selected, each data-analytic method provides the investigator with an array of well developed techniques suitable for analyzing family interaction data.
Moreover, advanced statistical packages containing these methods are now available for the micro-computer (e.g., SPSS-PC+, ( Nie, 1986)), as are specialized programs like the Williams and Gottman ( 1981) time series program, and Bakeman's ELAG ( 1983). As a result, the cost of calculating and the ease of interpreting the results of these approaches are within the capacity of most researchers. Clearly this chapter was written to illustrate these methods, but more importantly, it was written to encourage family interaction researchers to utilize these powerful methods as a means of analyzing behavioral patterns.
The authors wish to acknowledge the support for preparation of this manuscript by NIMH grant MH 18262-02 to the first author and by NIMH Research Career Development Award K00257 and NIMH Grant 1RO1MH42722-01 to the second author.
The editors gratefully acknowledge the assistance of David Baldwin in reviewing the more technical material in this chapter.