# Statistical Advisor, How To Test Predictions About Time-Dependent (Lagged) Relationships

TIME SERIES: This chapter discusses various facilities for describing and analyzing time-lagged relationships. For example, advertising expenditures may correlate with sales over time, however, with some lag (because it may take a while before the effect of a new advertisement is felt). A variable may also correlate with itself, with some lag. For example, retail sales may go through a regular 12 month cycle, with most sales in the month of December. Thus, if monthly sales were recorded in a file, every observation could correlate fairly highly with the 13th observation preceding it. Such correlations are referred to as autocorrelations.

Distributed Lags is a specialized option for analyzing the lagged effects of one independent (predictor, exogenous) variable with one dependent (criterion, endogenous) variable. Unlike the autocorrelation feature in Time Series, Distributed Lags can simultaneously evaluate multiple lags.

NEURAL NETWORKS: This chapter describes numerous options for building neural networks to predict lagged response variables.