There is an ambiguity of whether to regress y(x) or x(y), which typically give different results. One way to assess this is to see how correlated the residuals of the forward and backwards fits are with x: the correct regression direction produces the smaller residual correlation. This can be used in Roxy by plotting both residual correlations, calculating their Spearman and Pearson coefficients and showing them to a user with a recommendation for which variable to take as the independent one. Roxy should regress by default in this direction (although the user may stipulate which variable is independent if they have prior knowledge about this).
There is an ambiguity of whether to regress y(x) or x(y), which typically give different results. One way to assess this is to see how correlated the residuals of the forward and backwards fits are with x: the correct regression direction produces the smaller residual correlation. This can be used in Roxy by plotting both residual correlations, calculating their Spearman and Pearson coefficients and showing them to a user with a recommendation for which variable to take as the independent one. Roxy should regress by default in this direction (although the user may stipulate which variable is independent if they have prior knowledge about this).