Checking for Stationarity by Gregory N. Hight
A price chart is really a time series, and if you're using technical analysis on one segment, then it's important that the statistical characteristics are the same throughout the data. This is called stationarity. Here's a method for verifying that the data is statistically consistent.
It doesn’t matter whether you analyze markets using technical or fundamental methods. If you’re interested in trends and patterns, you’re using time series data.
You’ve probably illustrated trends and patterns using a chart with time on the horizontal axis and the
dependent measure (price, volume and so forth) on the vertical axis. This format is simply a time series format. So whenever a time series is discussed, think of a simple bar chart.
Logic favors the idea that all elements of a single time series — that is, the individual observations — share defining properties. Observations in the first half of a time series should measure the same thing as in the second half. What sort of a time series might not? Consider a time series composed of a single corporation’s stock price before and after a merger. A merger could be a radical change in the fundamentals of the company, and so the data would lack the homogeneity needed for all observations to share defining properties. Other, less conspicuous changes, such as shifts in industry, market, economy or political stability, can erode the necessary homogeneity. You can’t compare apples and oranges in the same time series.