seaborn: Cannot use dates on x-axis of lmplot

I would like to use lmplot to visualize polling data for multiple candidates as demonstrated in this notebook, but I get TypeError: invalid type promotion.

I can convert the dates to a Days integer, but then I don’t get the date-based axis labels I want. This kind of use of dates is handled transparently when I use R and ggplot2.

Here is the code inline:

import pandas as pd
import seaborn as sns
import StringIO

DATA = """Date,Candidate,Percent
2014-06-01,Smith,43
2014-06-01,Perkins,33
2014-06-05,Smith,44
2014-06-05,Perkins,31
2014-06-10,Smith,49
2014-06-10,Perkins,28
2014-06-12,Smith,49
2014-06-12,Perkins,25
2014-06-19,Smith,49
2014-06-19,Perkins,27"""
df = pd.read_csv(StringIO.StringIO(DATA), parse_dates=["Date"])

# This gives the TypeError
sns.lmplot("Date", "Percent", df, hue="Candidate")

# Converting the date objects to an integer works, but gives
# an innapropriately formatted table.
df["Day"] = df["Date"].map(lambda d: d.day)
sns.lmplot("Day", "Percent", df, hue="Candidate")

About this issue

Original URL
State: closed
Created 10 years ago
Reactions: 2
Comments: 18 (3 by maintainers)

Most upvoted comments

It is a well-established technique to use a lowess regression combined with a scatterplot to show election polling trend lines:

http://elections.huffingtonpost.com/pollster/2016-general-election-trump-vs-clinton

Is there a good way of doing this kind of plot in Seaborn? The naive approach doesn’t work:

import pandas as pd
import seaborn as sns

p = pd.read_csv("http://elections.huffingtonpost.com/pollster/api/charts/2016-general-election-trump-vs-clinton.csv", parse_dates=["start_date", "end_date"])
sns.regplot(["Clinton", "Trump"], "end_date", data=p, lowess=True)

bwbensonjr on Aug 24, 2016

+1 for support for pandas datetime64 on the x-axis of an lmplot. Converting the datetime64 to seconds-since-1970 (https://stackoverflow.com/questions/13703720/converting-between-datetime-timestamp-and-datetime64/13704307#13704307) lets me run the lmplot and gives me a useful model.

ianozsvald on Aug 25, 2015