有答案了参见软件作者的回复：
In any regression model you can test if the magnitude of explained sum
of squares differs from random expectation. While the R2 is the ratio of
explained / total sum of squares, the F is the ratio ofexplained /
residual variance. The F and P statistics thus measure the probability
of explaining as much (or more) of the response variable as the model
has explained if the null hypothesis were true.

Same for the local P-Value, which is the test of explanatory power of
the local regression performed for each cell.

请问你这个是怎么做出来的？