def ensembleMSDalltogether(filenames):
currentDir = os.getcwd()
sd_ensemble = []
dt_ensemble = []
for file1 in filenames:
dt_filename = currentDir + "/dt_" + file1
sd_filename = currentDir + "/sd_" + file1
if os.path.isfile(dt_filename) and os.path.isfile(sd_filename):
dt_data = np.load(dt_filename)
sd_data = np.load(sd_filename)
else:
stats, sd_data, dt_data = MSD.singleTrajectoryAnalysis(file1)
(i, _) = sd_data.shape
k = dt_data.size
assert (i == k), "sd and dt data mismatch in: " + file1
sd_list = arMa.makeListOfLists(sd_data)
dt_list = list(dt_data)
sd_ensemble, dt_ensemble = arMa.mergeLists(sd_ensemble, dt_ensemble,
sd_list, dt_list)
stats = np.zeros((len(dt_ensemble), 3)) # msd, std, dt
for a in range(len(dt_ensemble)):
stats[a, 2] = dt_ensemble[a]
stats[a, 0] = 3 * statistics.mean(sd_ensemble[a])
stats[a, 1] = 3 * statistics.stdev(sd_ensemble[a]) / math.sqrt(
len(sd_ensemble[a]))
# Oho, but what's this? We have some experimental error that makes the y-intercept non-zero!
# I tried assuming it's the same for every cell type, but that might be incorrect.
# We'll extrapolate it from a linear fit of the first 7 non-zero dts, then subtract it.
if dt_ensemble[0] == 0:
dt_slice = dt_ensemble[1:8]
statSlice = stats[1:8, 0]
linearFit = np.polyfit(dt_slice, statSlice, 1)
intercept = linearFit[0]
stats[1:, 0] = stats[1:, 0] - intercept
stats[1:, 1] = stats[1:, 1] + intercept
else:
dt_slice = dt_ensemble[0:7]
statSlice = stats[0:7, 0]
linearFit = np.polyfit(dt_slice, statSlice, 1)
intercept = linearFit[0]
stats[:, 0] = stats[:, 0] - intercept
stats[:, 1] = stats[:, 1] + intercept
# Stats: [dtIndex] [MSD | stddev | dt]
return stats
def ensembleMSD(filenames):
currentDir = os.getcwd()
stats_ensemble = []
dt_ensemble = []
# dt_mean_ensemble = []
# dt_std_ensemble = []
# means_ensemble = []
# stds_ensemble = []
for file1 in filenames:
stats_filename = currentDir + "/stats_" + file1
if os.path.isfile(stats_filename):
stats = np.load(stats_filename)
dt_data = stats[:, 2]
else:
stats, _, dt_data = MSD.singleTrajectoryAnalysis(file1)
lenDt = len(dt_data)
for dtIndex in range(lenDt):
stats_ensemble, dt_ensemble = arMa.dataInsert(
stats_ensemble, dt_ensemble, stats[dtIndex, :],
dt_data[dtIndex])
# means_ensemble, dt_mean_ensemble = arMa.dataInsert(means_ensemble, dt_mean_ensemble, stats[dtIndex,0], dt_data[dtIndex])
# stds_ensemble, dt_std_ensemble = arMa.dataInsert(stds_ensemble, dt_std_ensemble, stats[dtIndex,1], dt_data[dtIndex])
# So now, stats_ensemble is a list of lists of numpy arrays, and dt_ensemble is a list
# We will perform a weighted average of each MSD in the ensemble, with the weights equal to the inverse of the variance for each point
# assert (dt_mean_ensemble == dt_std_ensemble), "Something went wrong"
lenDt = len(dt_ensemble)
finalStats = np.zeros((lenDt, 3))
if dt_ensemble[0] == 0:
for a in range(1, lenDt):
finalStats[a, 2] = dt_ensemble[a]
thisMean, thisSD = weightedMean(stats_ensemble[a])
finalStats[
a,
0] = 3 * thisMean # Multiple of three comes from converting the radial MSD to the genomic MSD.
finalStats[a, 1] = 3 * thisSD
else:
for a in range(lenDt):
finalStats[a, 2] = dt_ensemble[a]
thisMean, thisSD = weightedMean(stats_ensemble[a])
finalStats[
a,
0] = 3 * thisMean # Multiple of three comes from converting the radial MSD to the genomic MSD.
finalStats[a, 1] = 3 * thisSD
# Oho, but what's this? We have some experimental error that makes the y-intercept non-zero!
# I tried assuming it's the same for every cell type, but that might be incorrect.
# We'll extrapolate it from a linear fit of the first 7 non-zero dts, then subtract it.
if dt_ensemble[0] == 0:
dt_slice = dt_ensemble[1:8]
statSlice = finalStats[1:8, 0]
linearFit = np.polyfit(dt_slice, statSlice, 1)
intercept = linearFit[0]
finalStats[1:, 0] = finalStats[1:, 0] - intercept
else:
dt_slice = dt_ensemble[0:7]
statSlice = finalStats[0:7, 0]
linearFit = np.polyfit(dt_slice, statSlice, 1)
intercept = linearFit[0]
finalStats[:, 0] = finalStats[:, 0] - intercept
return finalStats
