Making a regression line through a bar char using pandas or seaborn

Question

I am new to Pandas and Seaborn and trying to learn. I am trying to add a trend line and a bar plot on the same graph. I have some data that looks like

Year     Sample Size
 2000      500
 2001      3000
 2003      10000
 2004      20000
 2004      23000

I am new to pandas and seaborn and I am attempting to draw a line through the bar plot showing a decreasing or an increasing trend but struggling to do it on the same graph. Till now, I have a bar plot. Below you can find the code.

sampleSizes['Sample Size'] -> is the column I am plotting. It has about 12 values for 12 years.

plt.figure()
ax = sampleSizes['Sample Size'].plot(kind='bar', title="Trend of Sample Sizes", figsize=(15, 10), legend=True, color = 'grey', fontsize=8)
plt.show()

I am struggling to add a trend line to this. I would be grateful if someone could point me in the right direction.

UPDATE

FinancialYear  Sample Size
   2001         2338
   2002         3171
   2003         2597
   2004         2740
   2005         3447
   2006         3098
   2007         2610
   2008         2819
   2009         2057
   2010         2174
   2011         2709

enter code here

Is a bar plot the right type of representation for this? Or would a line plot be more appropriate? There's also [`seaborn.regplot`](https://seaborn.pydata.org/generated/seaborn.regplot.html), but it sounds like your data is not linear at all. — Arya McCarthy, Feb 17 '18 at 14:24
@AryaMcCarthy yes, I agree line plot would be better but I am trying to have the bar graphs and a line showing the trend of the data on the same graph. I hope it makes it clear what I'm trying to achieve. I have edited the data above a bit. I just randomly got these numbers. The numbers are not necessarily representative of my data. However, the variables are — paddy, Feb 17 '18 at 14:33

MaxU - stand with Ukraine · Accepted Answer · 2018-02-17T16:54:59.243

7

UPDATE2: using updated data set

In [250]: lr = Ridge()

In [251]: lr.fit(df[['FinancialYear']], df['Sample Size'])
Out[251]:
Ridge(alpha=1.0, copy_X=True, fit_intercept=True, max_iter=None,
   normalize=False, random_state=None, solver='auto', tol=0.001)

In [252]: plt.bar(df['FinancialYear'], df['Sample Size'])
Out[252]: <Container object of 11 artists>

In [253]: plt.plot(df['FinancialYear'], lr.coef_*df['FinancialYear']+lr.intercept_, color='orange')
Out[253]: [<matplotlib.lines.Line2D at 0x171def60>]

Result:

UPDATE:

In [184]: from sklearn.linear_model import Ridge

In [185]: lr = Ridge()

In [186]: lr.fit(df[['Year']], df['Sample Size'])
Out[186]:
Ridge(alpha=1.0, copy_X=True, fit_intercept=True, max_iter=None,
   normalize=False, random_state=None, solver='auto', tol=0.001)

In [187]: plt.bar(df['Year'], df['Sample Size'])
Out[187]: <Container object of 5 artists>

In [188]: plt.plot(df['Year'], lr.coef_*df['Year']+lr.intercept_, color='orange')
Out[188]: [<matplotlib.lines.Line2D at 0x17062898>]

Result:

Try to use matplotlib methods for that:

plt.bar(df['Year'], df['Sample Size'])
plt.plot(df['Year'], df['Sample Size'], '-o', color='orange')

Result:

edited Feb 17 '18 at 16:54

answered Feb 17 '18 at 15:30

MaxU - stand with Ukraine

205,989
36
386
419

I am able to do that. I am trying to make a smoother line like a regression line just showing the general trend. Like a straight upward line (like in a linear regression) instead of it being joined. – paddy Feb 17 '18 at 15:46
@paddy, please check updated answer - is that what you want? – MaxU - stand with Ukraine Feb 17 '18 at 16:01
OMG YES. Thank you so much. Been trying so hard for the past hour – paddy Feb 17 '18 at 16:12
@MaxU Its weird. So your code works when I have all 12 values. When I remove one because of it being an outlier, it seems to crash. Any reason you could see why that would happen? – paddy Feb 17 '18 at 16:44
@paddy, can you provide a reproducible data set? – MaxU - stand with Ukraine Feb 17 '18 at 16:45
@MaxU Updated my description. When I remove the value from year 2000, it crashes – paddy Feb 17 '18 at 16:49
@paddy, linear function doesn't fit well to this data. I'd think of using higher polynomial here... – MaxU - stand with Ukraine Feb 17 '18 at 16:56
@MaxU so what did you use here? It seems you've pasted the same code by mistake – paddy Feb 17 '18 at 17:02

Making a regression line through a bar char using pandas or seaborn

1 Answers1

Linked