>>> import numpy as np
>>> from scipy.stats import ttest_ind, ttest_rel
>>> results_model_0 = [0]*(6000 - 5876) + [1]*5876    # 0.9793 (accuracy for our best shallow model)
>>> results_model_1 = [0]*(6000 - 5910) + [1]*5910    # 0.9850 (accuracy for our best deep model)
>>> ttest_rel(results_model_1, results_model_0)    # same sort order; minimum standard error (largest test statistic)
Ttest_relResult(statistic=5.847056157218386, pvalue=5.267034245912735e-09)
>>> ttest_rel(results_model_1, results_model_0[::-1])    # opposite sort order; maximum standard error (smallest test statistic)
Ttest_relResult(statistic=2.3250453293684994, pvalue=0.020102730428109634)
>>> ttest_ind(results_model_1, results_model_0)
Ttest_indResult(statistic=2.3455397024038467, pvalue=0.0190157522891415)
>>> np.random.shuffle(results_model_0)    # random pairing
>>> np.random.shuffle(results_model_1)    # random pairing
>>> ttest_rel(results_model_1, results_model_0[::-1])
Ttest_relResult(statistic=2.3359967089478295, pvalue=0.019524182254221718)