The use of error formulas with time series
This chapter is concerned with universes which are extended over time and in which both the original observations and subsequent new ones are measured at successive intervals of time. The corn yield example of Chapter 14 and the steel cost example of Chapter 16 both used time-series data.
Many problems important in economics and other social sciences involve measurements in time. But time-series problems also arise in the biological and physical sciences. In recent years statisticians have recognized the similarities between time-series problems in the various disciplines and have made considerable progress toward their solution.
Differences Between Time-Series and Other Types of Data. The attitudes of research workers toward regression analysis of time series have varied between widely separated extremes. In the early and middle 1920's many researchers were completely unaware of problems connected with the sampling significance of time series. Then, under the (partly misinterpreted) influence of articles such as Yule's on "nonsense-correlations," it became fashionable to maintain that error formulas simply did not apply to time series.1 There was some implication that reputable statisticians should leave time series alone. But in some fields a large amount of data already existed in the form of time series; and the variables so recorded were frequently important in the theories of the____________________
G. U. Yule, "Why do we sometimes get nonsense-correlations between time-series?--A study in sampling and the nature of time series", Journal of the Royal Statistical Society, Vol. 89, No. 1, pp. 1-64, 1926.