如何通过循环基于已有DataFrame生成新DataFrame并合并？

Got it, here's a straightforward way to loop through generating those modified DataFrames and combine them all into one big dataset:

Approach #

• Start by storing all your DataFrames in a list, beginning with your original df.
• For each new DataFrame you want to create, take the most recent one from the list, apply your modification rules (multiply avgTemp by tempchange, add 20 to year), and add it to the list.
• Once all copies are generated, concatenate everything into a single DataFrame.

Code Implementation #

import pandas as pd

# Configure your parameters here
num_additional_dfs = 3  # Number of new DataFrames to generate (df2, df3, df4)
# tempchange = your existing temperature multiplier (e.g., ~1.07 from your sample)

# Initialize list with original DataFrame
df_list = [df]

for _ in range(num_additional_dfs):
    # Get the last modified DataFrame to base the next one on
    last_df = df_list[-1]
    # Create a copy to avoid altering the original
    new_df = last_df.copy()
    # Apply your calculation logic
    new_df['avgTemp'] = new_df['avgTemp'] * tempchange
    new_df['year'] = new_df['year'] + 20
    # Add the new DataFrame to our collection
    df_list.append(new_df)

# Combine all DataFrames into the final dataset
df_final = pd.concat(df_list)

How It Works #

Let's use your sample data to illustrate:

• Starting with df (year 2012), the first loop creates df2 (year 2032, avgTemp scaled by tempchange)
• The second iteration takes df2 to make df3 (year 2052, avgTemp scaled again)
• The third iteration generates df4 (year 2072, avgTemp scaled once more)
• The final df_final will have all these rows stacked, matching your sample output but with as many additional years as you specify.

Optional Variation #

If you want each new DataFrame to be based on the original df instead of the previous modified one (e.g., each year uses the original temp multiplied by tempchange raised to the power of the iteration count), just replace last_df with df inside the loop. But based on your question, building on the prior modified DataFrame is exactly what you need.

内容的提问来源于stack exchange，提问作者Timo