如何在Python中实现R语言rep()+filter()的滑动均值计算？

Got it, let's break down what your R code is doing first, then jump into the Python equivalents.

In R, filter(x, rep(1/30, 30)) computes a 30-element sliding average where each window's value is the mean of the current element and the previous 29 elements (with boundary handling for the first 29 values, where it uses a smaller window of available elements). The rep(1/30, 30) creates equal weights for each element in the window, which is exactly how a simple moving average works.

Here are the best ways to replicate this in Python:

Option 1: Use Pandas (Simplest, Matches R's Boundary Behavior) #

Pandas' rolling method is designed for exactly this kind of sliding window calculation, and it handles the edge cases (smaller initial windows) exactly like R's filter does:

import pandas as pd
import numpy as np

# Create the same array as R's 1:300
x = np.arange(1, 301)
# Convert to a pandas Series for rolling calculations
x_series = pd.Series(x)

# Calculate 30-window sliding average, with min_periods=1 to allow smaller initial windows
sliding_avg = x_series.rolling(window=30, min_periods=1).mean()

# Print the result (use .values to get a numpy array if needed)
print(sliding_avg.values)

This will give you a 300-element array, just like your R code—each value from index 0 to 28 is the mean of the first i+1 elements, and from index 29 onward it's the mean of the previous 30 elements.

Option 2: Use NumPy (No Pandas Dependency) #

If you can't use Pandas, you can use np.convolve, but you'll need to manually handle the boundary cases to match R's behavior (since np.convolve defaults to either trimming edges or padding with zeros, which doesn't match R's partial-window approach):

import numpy as np

x = np.arange(1, 301)
window_size = 30
weights = np.repeat(1 / window_size, window_size)

# Calculate the full-window averages (valid mode gives 300 - 30 + 1 = 271 values)
full_window_avg = np.convolve(x, weights, mode='valid')

# Calculate the partial-window averages for the first 29 elements
partial_window_avg = np.array([x[:i+1].mean() for i in range(window_size - 1)])

# Combine partial and full window results to get a 300-element array
sliding_avg = np.concatenate([partial_window_avg, full_window_avg])

print(sliding_avg)

If you don't care about the partial initial windows and just want the full 30-element averages (resulting in 271 values), you can skip the partial window part and just use np.convolve(x, weights, mode='valid').

A Quick Note on Matching R's Exact Output #

R's filter function defaults to side=1 (using past values), which is exactly what both of these Python approaches implement. If you ever need side=2 (symmetric window using past and future values), you can adjust the Pandas rolling method with center=True, or adjust the NumPy convolution mode accordingly.

内容的提问来源于stack exchange，提问作者scriptgirl_3000