Zillow says its estimates have a median error rate of 7.9percentthoughugh this also changes depending on the geographical area).
When testing for differences in the recognition of shoes, we observed higher error rates among CPs, median error rate of 24.1%, compared to controls, median of 18.4% (W = 223, n0 = 25, nCP = 13, p = 0.032 one-sided).
Testing for group differences in face recognition performance, we found that CPs, with a median error rate of 14.7%, perform significantly worse than controls, with a median of 6.3% (Wilcoxon rank sum test, W = 267, n0 = 25, nCP = 13, p<0.001 one-sided).
OOB classification errors are reported as the median error rate across all 1,000 replicates of each model.
This is significantly better than random guess, having a median error rate of 53%, and 95% confidence intervals of [32%, 74%].
In comparison, the error percentage for the CF scheme was higher in this round than the median error rate registered in previous rounds (0.2%) [ 14].
Since the error rate fit a Poisson distribution, we chose to use the Wilcoxon statistical test to compare the median error rates between the clinical and clonal samples and between the different genes for each samples.
In this case, we find median error rates of ∼0.1 0.2.
But still the median error rates are <17%.
The remaining algorithms had a similar performance, with median error rates roughly between 10 and 20percentt.
For all four structures, the estimation of P is very precise for ϵ=0.01, with median error rates < 0.05 (Fig. 2E).
