Statistical Tests of Indicator Effectiveness
Overview
A range indicator is one that can assume any value of a continuous set. For example, the
momentum indicator
can take a range of values. When this is the case, it is not possible to separate the data points into discrete sets, unless
one wants to set a threshold in the indicator (such as a momentum higher than 20% or something to that effecct).
Testing Range Indicators
The typical way to test an indicator that can take a range of values is to assume that the returns of the asset can take a range of
values that is proportional in some way to the indicators value. This is tested by running
a
regression
of the assets return against the indicators value.
That is, it is assumed that the following equation holds
{% r = \alpha + \beta \times indicator %}
Then a regression is run to determine if {% \beta %} has a non-zero value. Then the tools of
regression inference
can be used to create confidence intervals around the result.
