03.07

jar_functions.py

@@ -63,14 +63,14 @@ def mean_noise_cut(frequencies, time, n):
         f = np.mean(frequencies[k:k+n])
         cutf.append(f)
         cutt.append(t)
+
     return cutf, cutt
 
+
 def step_response(t, a1, a2, tau1, tau2):
-    r_step = a1*(1 - np.exp(-t/tau1)) + a2*(1- np.exp(-t/tau2))
+    r_step = (a1*(1 - np.exp(-t/tau1))) + (a2*(1 - np.exp(-t/tau2)))
+    r_step[t<0] = 0
     return r_step
-# plotten mit manual values for a1, ...
-# auch mal a1 oder a2 auf Null setzen.
-
 
 
 def base_eod(frequencies, time, onset_point):
@@ -82,6 +82,7 @@ def base_eod(frequencies, time, onset_point):
     base_eod.append(base)
     return base_eod
 
+
 def JAR_eod(frequencies, time, offset_point):
     jar_eod = []
 
@@ -90,4 +91,35 @@ def JAR_eod(frequencies, time, offset_point):
     jar = np.mean(frequencies[(time >= offset_start) & (time < offset_point)])
     jar_eod.append(jar)
 
-    return jar_eod
+    return jar_eod
+
+
+def mean_loops(start, stop, timespan, frequencies, time):
+    minimumt = min(len(time[0]), len(time[1]))
+    # new time with wished timespan because it varies for different loops
+    tnew = np.arange(start, stop, timespan / minimumt)  # 3rd input is stepspacing:
+                                                        # in case complete measuring time devided by total number of datapoints
+    # interpolation
+    f0 = np.interp(tnew, time[0], frequencies[0])
+    f1 = np.interp(tnew, time[1], frequencies[1])
+
+    #new array with frequencies of both loops as two lists put together
+    frequency = np.array([f0, f1])
+
+    #making a mean over both loops with the axis 0 (=averaged in y direction, axis=1 would be over x axis)
+    mf = np.mean(frequency, axis=0)
+    return mf, tnew
+
+
+def norm_function(cf_arr, ct_arr, onset_point, offset_point):
+    onset_end = onset_point - 10
+    offset_start = offset_point - 10
+
+    base = np.mean(cf_arr[(ct_arr >= onset_end) & (ct_arr < onset_point)])
+
+    ground = cf_arr - base
+
+    jar = np.mean(cf_arr[(ct_arr >= offset_start) & (ct_arr < offset_point)])
+
+    norm = ground / jar
+    return norm

second_try.py

@@ -4,11 +4,14 @@ import glob
 import IPython
 import numpy as np
 from IPython import embed
+from scipy.optimize import curve_fit
 from jar_functions import parse_dataset
 from jar_functions import mean_noise_cut
 from jar_functions import step_response
 from jar_functions import JAR_eod
 from jar_functions import base_eod
+from jar_functions import mean_loops
+from jar_functions import norm_function
 
 
 datasets = [(os.path.join('D:\\jar_project\\JAR\\2020-06-22-ac\\beats-eod.dat'))]
@@ -27,58 +30,30 @@ timespan = 210
 
 for dataset in datasets:
     #input of the function
-    t, f, a, e, d, s= parse_dataset(dataset)
-
-    minimumt = min(len(t[0]), len(t[1]))
-    # new time with wished timespan because it varies for different loops
-    tnew = np.arange(start, stop, timespan / minimumt)  # 3rd input is stepspacing:
-                                                        # in case complete measuring time devided by total number of datapoints
-    # interpolation
-    f0 = np.interp(tnew, t[0], f[0])
-    f1 = np.interp(tnew, t[1], f[1])
-
-    #new array with frequencies of both loops as two lists put together
-    frequency = np.array([f0, f1])
-
-    #making a mean over both loops with the axis 0 (=averaged in y direction, axis=1 would be over x axis)
-    mf = np.mean(frequency, axis=0)
-
-    #appending data
-    eodf.append(e)
-    deltaf.append(d)
-    stimulusf.append(s)
-    amplitude.append(a)
-    frequency_mean.append(mf)
-    time.append(tnew)
-
-"""
-    for a in [0, 1, 2]:
-        for b in [0, 1, 2]:
-            r_step = step_response(t = ct_arr, a1 = a, a2 = b, tau1 = 30, tau2 = 60)
-"""
-
-for i in range(len(frequency_mean)):
-    for n in [100, 500, 1000]:
-        cf, ct = mean_noise_cut(frequency_mean[i], time[i], n=n)
+    t, f, a, e, d, s = parse_dataset(dataset)
+    mf , tnew = mean_loops(start, stop, timespan, f, t)
+embed()
 
+for i in range(len(mf)):
+    for n in [500, 1000, 1500]:
+        cf, ct = mean_noise_cut(mf[i], time[i], n=n)
 
-        ct_arr = np.array(ct)
         cf_arr = np.array(cf)
+        ct_arr = np.array(ct)
 
-        base = base_eod(cf_arr, ct_arr, onset_point = 0)
-        ground = cf_arr - base
-        jar = JAR_eod(ground, ct_arr, offset_point = 100)
-        norm = ground / jar
+        norm = norm_function(cf_arr, ct_arr, onset_point = 0, offset_point = 100)
 
         plt.plot(ct_arr, norm, label='n=%d' % n)
 
+        #r_step = step_response(t=ct_arr, a1=0.58, a2=0.47, tau1=11.7, tau2=60)
+
+        #plt.plot(ct_arr[ct_arr < 100], r_step[ct_arr < 100], label='fit: n=%d' % n)
 
-for n in [1480]:
-    cf, ct = mean_noise_cut(frequency_mean[i], time[i], n=n)
-    ct_arr = np.array(ct)
-    cf_arr = np.array(cf)
-    r_step = step_response(t=ct_arr + 10, a1=0.55, a2=0.89, tau1=11.2, tau2= 280)
-    plt.plot(r_step, label='fit: n=%d' % n)
+        step_values, step_cov = curve_fit(step_response, ct_arr[ct_arr < 100], norm [ct_arr < 100])
+
+        plt.plot(ct_arr [ct_arr < 100], step_response(ct_arr, *step_values)[ct_arr < 100], 'r-',  label='fit: a1=%.2f, a2=%.2f, tau1=%.2f, tau2=%.2f' % tuple(step_values))
+        print(step_values)
+const_line = plt.axhline(y=0.632)
 
 'plotting'
 plt.xlim([-10,220])
@@ -87,7 +62,14 @@ plt.xlabel('time [s]')
 plt.ylabel('rel. JAR magnitude')
 #plt.title('fit_function(a1=0)')
 #plt.savefig('fit_function(a1=0)')
-plt.legend()
+plt.legend(loc = 'lower right')
 plt.show()
 embed()
-# Zeitkonstante: von sec. 0 bis 63%? relative JAR
+
+# noch mehr in funktionen reinhauen (quasi nur noch plotting und funktionen einlesen)
+# zeitkonstanten nach groß und klein sortieren
+# onset dauer auslesen
+# ID aus info.dat auslesen
+# alle daten einlesen durch große for schleife (auch average über alle fische?)
+# für einzelne fische fit kontrollieren
+