Before discussing selection that would bias assessments of special language program performance downward, I will address two factors that could
introduce upward biases.
Not exact matches
What our analysis does do is confirm that the explicitly stated data issues that come with the JTWC best track do have a temporal dependence in them, and as it happens, this
bias introduces an
upward trend in metrics based on intensity.
If the
upward jumps are greater in number than the downward jumps, but they have equal means and standard deviations, then I don't see how the algorithm
introduces a
bias.