python matplotlib confidence interval boxes

Question

can someone please guide me to produce a plot with python like the one in this picture: confidence boxes

I am measuring the runtime of a process depending on one parameter which has 18 different values. I made 100 runs of measuring the runtime for each value. Now I need the 99%(or 95%) confidence interval as boxes around the mean for each value of the parameter.

I thought about using seaborn.tsplot() with interpolate=False and then tweaking the matplot output to produce confidence boxes like that

Here is a minimal python example of what the pandas dataFrame looks like

import pandas as pd
import numpy as np
import seaborn as sns
import scipy.stats as st

RUNS=100

X = list(range(20,210,10))

data = pd.DataFrame()
for x in X:
    fakeData = np.array([x*10]*RUNS)
    randomVariation = np.random.random(RUNS)*x
    y = np.add(fakeData, randomVariation)
    data[x] = y

#calculate mean and confidence intervals
for x in X:
    y = np.array(data[x])
    mean = np.mean(y)
    ci = st.t.interval(0.99, len(y)-1, loc=mean, scale=st.sem(y))
    print((mean,)+ ci)

EDIT: forgot to mention: I already can calculate the mean and confidence intervals. I just need some guidance to do the graphical part.

Btw, `[i for i in range(20,210,10)]` can be written as `list(range(20,210,10))` — couragewolf, Aug 14 '17 at 09:37
Thanks for your comments, I edited the description. @Y0da: I already use scipy to calculate the confidence intervals — obraunsdorf, Aug 14 '17 at 10:02
Have you looked at this: https://stackoverflow.com/questions/36988428/confidence-interval-for-a-seaborn-boxplot?rq=1 — tupui, Aug 14 '17 at 10:15
Have you seen this example on the matplotlib gallery? https://matplotlib.org/examples/statistics/errorbars_and_boxes.html — tmdavison, Aug 14 '17 at 10:24

python matplotlib confidence interval boxes

0 Answers0