Product and Process Comparisons
|Detecting trends by plotting the data points to see if a line with an obviously non-zero slope fits the points
Detecting trends is equivalent to comparing the process values to
what we would expect a series of numbers to look like if there were
no trends. If we see a significant departure from a model where
the next observation is equally likely to go up or down, then we
would reject the hypothesis of "no trend".
A common way of investigating for trends is to fit a straight line to the data and observe the line's direction (or slope). If the line looks horizontal, then there is no evidence of a trend; otherwise there is. Formally, this is done by testing whether the slope of the line is significantly different from zero. The methodology for this is covered in Chapter 4.
|Other trend tests
|A non-parametric approach for detecting significant trends known as the Reverse Arrangement Test is described in Chapter 8.