Rolling and keep the threshold

Question

I have a database in pandas with a date and a variable. I need to compare the values of the variable with its next three days and if any of its values are increased with respect to the initial value, copy the date in a new dataframe. However, I need that the threshold that is taken into account is that of the first group or the first row and only if the value is greater than this threshold, its value is changed by the new threshold.

Below is an example:

index	Date	variable	Threshold
1	2020-04-01	-1.128146	Initial threshold
2	2020-04-02	0.389000	After first iteration-New threshold
3	2020-04-03	-0.400516
4	2020-04-04	-0.077735
5	2020-04-05	0.5	After second iteration New threshold
6	2020-04-06	0.45
7	2020-04-07	0.25
8	2020-04-07	-0.3	After third iteration Keep before threshold (0.5)
9	2020-04-07	-0.2

Expect dataframe:

index	Date	variable
2	2020-04-02	0.389000
5	2020-04-05	0.5

Answer 1

Given the following toy dataframe:

import pandas as pd

df = pd.DataFrame(
    {
        "date": {
            1: "2020-04-01",
            2: "2020-04-02",
            3: "2020-04-03",
            4: "2020-04-04",
            5: "2020-04-05",
            6: "2020-04-06",
            7: "2020-04-07",
            8: "2020-04-07",
            9: "2020-04-07",
        },
        "variable": {
            1: -1.128146,
            2: 0.389,
            3: -0.400516,
            4: -0.077735,
            5: 0.5,
            6: 0.45,
            7: 0.25,
            8: -0.3,
            9: -0.2,
        },
    }
)

Here is one way to do it:

threshold = df.loc[1, "variable"]
indices = []
for idx in range(2, df.shape[0], 3):
    if (new_threshold:=max(df.loc[idx: idx+2, "variable"])) > threshold:
        threshold = new_threshold
        indices.append(df.loc[df["variable"] == new_threshold, "variable"].index[0])

new_df = df.loc[indices, :]

print(new_df)
# Output
         date  variable
2  2020-04-02     0.389
5  2020-04-05     0.500