Correlation is arguably the most important statistical metric in life sciences [Dolph book citation]. The Pearson coefficient rho ranges from -1 to +1, where -1 means perfect negative correlation, +1 perfect positive correlation, and 0 means no linear correlation whatosever.
In this step-by-step tutorial, learn how to create a correlation matrix for exploratory data analysis in Python from scratch.
The first step is building the matrix strucutre that ensures every feature is compared (and correlated with a Pearson coefficient) with every other feature at least once. Let's say our study has 5 variables or "features", and if each feature is to be compared with each other at least once, we will have 5^2 combinations in a matrix that is 5 rows tall and 5 columns long, with a total of 25 plots. We use the matplotlib.pyplot library (as plt) to create and render this matrix.
num_var = 5
fig = plt.figure(figsize=(9, 9))
for i in range(num_var ** 2):
ax = fig.add_subplot(num_var, num_var, i + 1)
ax.text(0.5, 0.5, str(i + 1), fontsize=20, ha="center", va="center")
plt.show()
def width_constrained_text(string_input, max_characters_per_line):
count = 0
lines_array = []
segments = string_input.split(" ")
for segment in segments:
if len(segment) < max_characters_per_line:
segment_plus = ""
required_length = 0
if count == 0:
required_length = len(segment)
segment_plus = segment
else:
required_length = len(segment) + 1
segment_plus = " " + segment
if count + required_length < max_characters_per_line:
if count == 0:
lines_array.append(segment_plus)
else:
lines_array[-1] += segment_plus
count += required_length
else:
count = len(segment)
lines_array.append(segment)
else:
lines_array.append(segment)
return "\n".join(lines_array)