For each pair of input-output parameters, DesignXplorer computes different kinds of correlation:
Pearson Correlation (linear relationship)
Spearman Correlation (monotonic relationship)
Quadratic Correlation (quadratic relationship)
For each correlation value, DesignXplorer computes the p-value of this correlation.
The p-value of a correlation value allows DesignXplorer to quantify the relevance of the correlation.
The following notation is used in equations:
the number of sample points
the observed correlation between
(an input parameter) and
(an output parameter)
the real correlation value between
and
the null hypothesis for
:
The test of the null hypothesis is:
The variable is approximately distributed like a Student's Distribution
where
degrees of freedom.
The p-value corresponds to the probability of obtaining the same assuming that the null hypothesis is true. The p-value is given by
.
When trying to find a relationship between and
, two types of errors are possible:
Type I Error (false positive): You think that there is
a relationship between and
, when there is in fact no relation between
and
.
Type II Error (false negative): You think that there is
no relationship between and
, when there is in fact a relationship between
and
.

The p-value corresponds to the value.
The larger the correlation value , the less likely it is that the null hypothesis is true.
The relevance value of the correlation value exposed at the DesignXplorer level is a transformation of the p-value from [0; 1] to [0; 1]:

When running a filtering method with a relevance threshold equal to 0.5 on the correlation value only, the major input parameters selected must be at least a p-value of the correlation coefficient with any filtering output parameter equals to or less than 0.05.