fix performance issue in t-test

WPI_SCA_LIBRARY/Metrics.py

@@ -126,6 +126,8 @@ def t_test_tvla(fixed_t: np.ndarray, random_t: np.ndarray, visualize: bool = Fal
     new_sf_outer = []
     new_sr_outer = []
     t_max_outer = []
+    fixed_t = np.array(fixed_t, dtype=np.float64)
+    random_t = np.array(random_t, dtype=np.float64)
 
     if len(fixed_t) != len(random_t):
         raise ValueError("Length of fixed_t and random_t must be equal")
@@ -151,12 +153,13 @@ def t_test_intermediate(mf_old, mr_old, sf_old, sr_old, new_tf, new_tr, n):
             new_sf = sf_old + (new_tf - mf_old) * (new_tf - new_mf)
             new_sr = sr_old + (new_tr - mr_old) * (new_tr - new_mr)
 
-            new_stdf = np.sqrt(np.array(new_sf / n, dtype=np.float64))
-            new_stdr = np.sqrt(np.array(new_sr / n, dtype=np.float64))
+            new_stdf = np.sqrt(new_sf / n)
+            new_stdr = np.sqrt(new_sr / n)
 
             try:
-                welsh_t = np.array(new_mr - new_mf) / np.sqrt(
-                    np.array((new_stdr ** 2)) / (n + 1) + np.array((new_stdf ** 2)) / (n + 1))
+                with np.errstate(divide='ignore'):
+                    welsh_t = np.array(new_mr - new_mf) / np.sqrt(
+                        np.array((new_stdr ** 2)) / (n + 1) + np.array((new_stdf ** 2)) / (n + 1))
             except ZeroDivisionError:
                 welsh_t = 0