我正在尝试使用scipy.t.sf来计算双样本T检验的p值。我知道最简单的方法是使用scipy.stats.ttest_ind，但我希望加深我的理解，看看数学并想出替代方法。
我的原始代码是：

import scipy.stats as st
test1 = (8,7,10,5,7)
test2 = (9,5,12,8)
result = st.ttest_ind(test1, #men's sample data
                       test2, #women's sample data
                       alternative = 'less', #alternative hypothesis is that women's waiting time is longer than men's
                       equal_var=False) #perform Welch’s t-test, which does not assume equal population variance.
pvalue = result.pvalue

字符串
输出为Ttest_indResult（统计值=-0.6641304531560304，p值=0.2684710269367842）
然而，当我尝试使用scipy.t.sf求解p值时，我无法得到0.26847的统计值。

test_1 = np.array(test1)
test_2 = np.array(test2) 
test_mean = test_1.mean()
ctrl_mean = test_2.mean()
mean_diff = test_mean - ctrl_mean
test_var = test_1.var()/(test_1.shape[0]-1)
ctrl_var = test_2.var()/(test_2.shape[0]-1)

tt = mean_diff/math.sqrt(test_var + ctrl_var) # (test_mean - ctrl_mean) / sqrt(var(test_mean - ctrl_mean))
n = min(test_1.shape[0],test_2.shape[0])
pval = st.t.sf(np.abs(tt), n-1)

型
现在的输出是（t =-0.6641304531560304，pvalue = 0.2770449975228821）。正如你所看到的，t与我们从scipy.stats.ttest_ind方法中观察到的一致;然而，pvalue不是。我可以对代码做任何修改来获得0.26847的数字吗？谢谢！！