when plotting autocorrelation function plot getting UFuncTypeError

when plotting autocorrelation function plot gettingUFuncTypeError: Cannot cast ufunc 'divide' output from dtype('float64') to dtype('int32') the error happens after the comment # Plot the autocorrelation function for X and Y. do I need to convert trace x and y to float?

import numpy as np
import matplotlib.pyplot as plt
# Define the probability matrix
p = np.array([[10/66, 15/66, 3/66], [20/66, 12/66, 0], [6/66, 0, 0]])
# Initialize starting values for X and Y
x = 0
y = 0
# Create arrays to store trace plots
trace_x = []
trace_y = []
# Run 12,000 iterations of Gibbs sampling (with 2,000 burn-in)
for i in range(12000):
    # Sample a new value for X given the current value of Y
    py = np.sum(p[y])
    x = np.random.choice([0, 1, 2], p=p[y] / py)
    # Sample a new value for Y given the current value of X
    px = np.sum(p[:, x])
    y = np.random.choice([0, 1, 2], p=p[:, x] / px)
    # Append the new values to the trace plots after the burn-in period
    if i >= 2000:
        trace_x.append(x)
        trace_y.append(y)
# Plot the autocorrelation function for X and Y
plt.acorr(trace_x, maxlags=50)
plt.title('Autocorrelation plot for X')
plt.xlabel('Lag')
plt.ylabel('Autocorrelation')
plt.show()
plt.acorr(trace_y, maxlags=50)
plt.title('Autocorrelation plot for Y')
plt.xlabel('Lag')
plt.ylabel('Autocorrelation')
plt.show()

A look into the traceback reveals that the error is due to the lines

    if normed:
        correls /= np.sqrt(np.dot(x, x) * np.dot(y, y))

inside the matplotlib source code at https://github.com/matplotlib/matplotlib/blob/main/lib/matplotlib/axes/_axes.py.

The error appears due to the prevention of automatic upcasting within the numpy library when using in-place operations. This has been beautifully explained in an answer by SO user: poke here. Now, to answer the OP's question, we need to steer clear of the in-place operation.

So, for the OP's specific problem, there are three solutions:

Solution 1. If normalised values are not needed, the normed=False can be passed in the plt.acorr function as an argument as shown below:

    plt.acorr(trace_x, maxlags=50, normed=False)

Output:

Solution 2. If normalised values are needed, cast the input to float first:

plt.acorr(np.array(trace_x, dtype=float), maxlags=50)

Output:

Solution 3. The fix can be applied inside the matplotlib's code itself. To change the in-place operator to regular assignment as:

correls = correls / np.sqrt(np.dot(x, x) * np.dot(y, y))

Output (Same as solution 2):

I put a pull request for this to matplotlib here and referred to the OP's question. I will update the answer if my pull-request is approved.

EDIT:
The pull request has been approved. The acorr function works as expected even when using a list or numpy array with all integers.