This tool will perform a paired-sample t-test to evaluate whether a significant statistical difference exists between the two rasters. The null hypothesis is that the difference between the paired population means is equal to zero. The paired-samples t-test makes an assumption that the differences between related samples follows a Gaussian distribution. The tool will output a cumulative probability distribution, with a fitted Gaussian, to help users evaluate whether this assumption is violated by the data. If this is the case, the wilcoxon_signed_rank_test should be used instead.
The user must specify the name of the two input raster images (input1
and input2
) and the output report HTML file (output
). The test can be performed optionally on the entire image or on a random sub-sample of pixel values of a user-specified size (num_samples
). In evaluating the significance of the test, it is important to keep in mind that given a sufficiently large sample, extremely small and non-notable differences can be found to be statistically significant. Furthermore statistical significance says nothing about the practical significance of a difference.
two_sample_ks_test, wilcoxon_signed_rank_test
def paired_sample_t_test(self, raster1: Raster, raster2: Raster, output_html_file: str, num_samples: int) -> None: ...