NSSDNResult.r_squared#
- property NSSDNResult.r_squared: float[source]#
One minus the one-step residual share of variance over the effective sample.
\[R^2 = 1 - \frac{\sum_{t=1}^{n-1} e_t^2} {\sum_{t=1}^{n-1} (y_t - \bar y)^2},\]with \(\bar y\) the mean of
endog[1:]. Zero when the target has no variance. A persistent series gives a random walk an \(R^2\) near one as well, so read this next to a naive benchmark rather than on its own.