[FIXED] Finding Multiple Matches Using Pandas Rolling Min

Issue

I am able to get the rolling minima of a column using:

df['MemberLows'] = df['Members'].rolling(5, closed="left").min()

Is there a way of returning values that are within a tolerance of the minima without resorting to a non-vectorized approach like a for loop?

For example, given the table:

Month	Members
Jan	10
Feb	10
Mar	10
Apr	6
May	5
Jun	10
Jul	10
Aug	10
Sep	10
Oct	7
Nov	8
Dec	10

The rolling minima would be 5 and 7. Say I would like to also return values within 20% of the minima; this would also return 6 and 8.

Ultimately I want to return two arrays. One containing the indices of the minima/those within 20% and one containing the actual values:

indices_arr = [3 4 9 10]

values_arr = [6 5 7 8]

Solution

You are almost there. Just calculate the relevant threshold for each value and compare. Like so:

>>> import pandas as pd
>>> df = pd.read_csv("table.tsv", delimiter="\t")
>>> result = df[df["Members"] <= (df['Members'].rolling(5, center=True, closed="left").min() * 1.2)]
>>> print(result)
   Month   Members
3    Apr         6
4    May         5
9    Oct         7
10   Nov         8

Two things worth noting:

• I'm assuming that values are positive (otherwise I can't just multiply by 1.2)
• I added center=True to get min value indices to correspond to indices of members.

Answered By - VRehnberg