Odtnhj

I am trying to write a program to determine the slope and intercept of a linear regression model over a moving window of points, i.e. from (x1, y1) to (x2, y2) and then from (x2, y2) to (x3, y3). I have successfully carried out a linear regression across the two numpy arrays (x and y), but I am not sure how to approach this project. I would like the window size to be a user-input parameter. I can reshape my two arrays using array subsetting and achieve the a window over which the linear regression is carried out, but i do not know how to automate this and how to save each slope and intercept into a file. I have tried my best, but I am a new programmer and don't know where to look. Can someone point me in the right direction? Thank you!

Below is my code:

import matplotlib.pyplot as plt
import seaborn as sns
import scipy as sp
import numpy as np
import math as math

##import data from file
data = np.genfromtxt('CA_data.csv', delimiter=',')
print(np.shape(data))
#print(data)
##subset 2D array into 1D arrays
Tinv = data[:,0]
A_data = data[:,1]
B_data = data[:,2]
C_data = data[:,3]



a = np.empty_like()
b = np.empty_like()

##linear regression
###for i,j in a.range(Tinv[i:j]):
 Tnew[i] = Tinv[i:j]




a, b = np.polyfit(Tinv, A_data,1)

print("slope = ", a)
print("intercept = ", b)

# #visualize optimized slope
# a_vals = np.linspace(-6500, 0, 400)
# rss = np.empty_like(a_vals)
# for i, a in enumerate(a_vals):
# rss[i] = np.sum((A_data - a*Tinv - b)**2)
# _ = plt.plot(a_vals, rss, '-')
# _ = plt.xlabel('slope')
# _ = plt.ylabel('residual sum of squares')
# _ = plt.show()

##Theoretical plot for optimal values
startx = Tinv[0]
endx = Tinv[-1]

#print(startx)
#print(endx)
x = np.array([startx, endx])
y = a*x + b

##plot the data
plt.plot(Tinv, A_data, marker = '.', color = 'r', linestyle = 'none')
plt.plot(x,y) #plot theoretical data
plt.margins(0.02)
plt.axis([0.00275, 0.0035, 0.5, 3.0])
plt.xlabel("x")
plt.ylabel('y')
plt.title('title')
plt.show()


# print(Tinv)
# print(A_data)
# print(B_data)
# print(C_data)

1. Wrap the modeling and plotting in a function.




2. Then call this function from another function that subsets the arrays to the user specified range before feeding the "cleaned" data to the prediction function.

pseudocode:

def select_window(data, start, stop): 
 clean data = data[start:stop,:]
 X = clean_data[without_y]
 y = clean_data[y]

 prediction_function(X, y)

def prediction_function(X,y):
 predict_on_X_and_y_and_plot_as_you_desire

Here's what I ended up with:

data = np.genfromtxt('data.csv', delimiter=',')

w = data[:,0]
x = data[:,1]
y = data[:,2]
z = data[:,3]


lenw = int(len(w))
lenx = int(len(x)) #change the value within inner parenthesis to suit different dataset

window = int(6) #change window size
wdata_avg = np.zeros(lenw - window + 1)
a = np.zeros(lenw-window+1)
b = np.zeros(lenx-window + 1)

for i in np.arange(len(w)):
 wdata = w[i:i + window]
 xdata = x[i:i+window]
 a[i], b[i] = np.polyfit(wdata, xdata,1)
 wdata_avg[i] = np.mean(wdata)
 if i == (lenw - window):
 break

There may be some inconsistencies in the code, since I tried to format it so it was general rather than specific to my data.