prepared pipeline

Random/graph.py

@@ -33,6 +33,8 @@ def ecdf_graph(N,RESOLUTION,seed,data):
         xlabel= "Value"
         ylabel="Cumulative Probability"
         title_name= name +"\n N = "+str(N)+"\n seed = "+str(seed) 
+        if N==0:
+            title_name= name +"\n seed = "+str(seed) 
         graph=sns.displot(data[name], kind="ecdf")
         graph.set(title=title_name)
         graph.set_axis_labels(xlabel,ylabel)

Random/main.py

@@ -57,9 +57,9 @@ def write_stats(data, name):
     skew      = scipy.stats.skew(data)
     exckurt   = scipy.stats.kurtosis(data) - 3
     #write file
-    with open(name+'.csv', 'w', newline='') as csvfile:
+    with open('stats.csv', 'a', newline='') as csvfile:
         datawriter = csv.DictWriter(csvfile, fieldnames=fields)
-        datawriter.writeheader()
+        #datawriter.writeheader()
         datawriter.writerow({
                     "name" : name,
                     "quantil25" : str(quantil25),
@@ -71,10 +71,26 @@ def write_stats(data, name):
                     "exckurt"   : str(exckurt)})
     return
 
+def write_stats_chi(data, name, RES):
+    fields=["chi","p"] 
+    val = stats.chisquared_uniform(data,RES)
+    chi = val[0]
+    p = val[1] 
+    #write file
+    with open('chi.csv', 'a', newline='') as csvfile:
+        datawriter = csv.DictWriter(csvfile, fieldnames=fields)
+        #datawriter.writeheader()
+        datawriter.writerow({
+                    "name" : name,
+                    "chi" : str(chi),
+                    "p" : str(p)})
+    return
+
 if __name__=="__main__":
     ##INIT PARAMS##
-    N = 100
-    RESOLUTION = 1000
+    N = 1000000
+    RESOLUTION = 100
+    iterN = [100, 1000000, 100000000] 
     
     #init generators
     generators = [ParkMiller(), KnuthLewis(), Marsaglia()
@@ -87,7 +103,7 @@ if __name__=="__main__":
     #set seaborn params
     sns.set_theme(style="darkgrid")
      
-    #generate all diagrams
+    #generate N random values for each RNG
     data = values(N,seed,generators)
     
     #graph.hist_distributivity_graph(N,RESOLUTION,seed,data)
@@ -96,8 +112,25 @@ if __name__=="__main__":
     
     #graph.ecdf_graph(N,RESOLUTION,seed,data)
     
-    graph.compare(0, RESOLUTION, seed
-                  , iteratives([1000, 10, 100], seed, ParkMiller()))
-    print(stats.chisquared_uniform(data["ParkMiller"],N))
-    print(np.quantile(data["ParkMiller"],0.25))
-    write_stats(data["ParkMiller"],"ParkMiller")
+    #graph.compare(0, RESOLUTION, seed
+    #              , iteratives([1000, 10, 100], seed, ParkMiller()))
+    #print(stats.chisquared_uniform(data["ParkMiller"],RESOLUTION))
+    #print(np.quantile(data["ParkMiller"],0.25))
+    #write_stats(data["ParkMiller"],"ParkMiller")
+    #write_stats_chi(data["ParkMiller"], "ParkMiller", RESOLUTION)
+
+    #generate a lot of diagrams
+    graph.hist_distributivity_graph(N,RESOLUTION,seed,data)
+    
+    for name in data:
+        write_stats(data[name],name)
+        write_stats_chi(data[name],name)
+
+    for name in generators:
+        data_iter = iteratives(iterN, seed, name)
+        
+        graph.ecdf_graph(0, RESOLUTION, seed, data_iter)
+        graph.compare(0, RESOLUTION, seed, data_iter)
+        
+        
+

Random/stats.py

@@ -5,15 +5,10 @@ Module to calculate various value on an array
 import numpy as np
 import scipy
 
-def chisquared_uniform(arr, N):
+def chisquared_uniform(arr,RES):
     """
     chisquared for uniform distrib
     """
-    _,counts = np.unique(arr, return_counts=True)
-    observed_freq = counts/N
-    #print(observed_freq)
-    expected_freq = np.full(shape=observed_freq.size,fill_value=1/N)
-    print(observed_freq),print(expected_freq)
-    print(np.sum(expected_freq), np.sum(observed_freq))
-    return scipy.stats.chisquare(observed_freq, expected_freq)
+    observed_freq, _  = np.histogram(arr, bins=np.linspace(0,1,RES))
+    return scipy.stats.chisquare(observed_freq)