Statistical researchers today enjoy many advantages over their counterparts of even a decade ago. Computers are more powerful and large datasets more available. Software for analysis and charting has also become more powerful and easier to use.
There is a lot of irony in these developments. The software has become so easy to "use" that the researcher no longer needs to understand basic principles to get results.
In the investment world, this takes the form of scores of misleading systems, charts, and tables, all purporting to show a relationship that does not really exist. The process of extracting information from data, or data mining, requires great skill if the results are to be meaningful. Since the untrained analyst takes all of the data and examines thousands of different possible relationships, something will always turn up. Statisticians call this "data dredging".
Sometimes it is not obvious that hundreds or thousands of combinations have been tested. This is particularly true when the result is a graph showing what appears to be a compelling relationship. In the investment world, these graphs appear all of the time.
Reader "RB" offered an interesting comment about the relationship between homebuilder sentiment and the stock market, pointing us to a chart showing the relationship. Take a look at the article at Calculated Risk, a site with plenty of good information on housing. The author does a good job of showing how looking at more data dramatically changes the picture.
Frequently, even if the relationship holds up over time, it is spurious. This is a statistical term meaning that both of the variables shown are the result of some other cause. In this case, one might hypothesize that something good (the economy perhaps?) led to similar action in both homebuilder expectations and stocks. By adjusting the scales and the lead time, the relationship can be made to appear even stronger.
Regular readers may remember that we did a similar analysis last September. Take a look for another example and further explanation.