Back to Cheatsheets

Table of Contents

1. ABOUT

• Import
• Show
• Close

2. FORMATTING

• Labels
• Legend
• Set Axis
• Modify Ticks
• Subplot
• Figure
• Seaborn Color

3. LINE GRAPHS

• Multiple
• Line Format
• Shade Error

4. BAR GRAPHS

• Multiple
• Stacked
• Error Bars

5. PIE CHART

6. HISTOGRAM

• Multiple

7. SCATTER

• 3D

8. BOXPLOT

ABOUT

• used to create graphs and charts
• if x_values are strings then x_values = range(len(string_list))

Import

from matplotlib import pyplot as plt

Show

• plt.show()

Close

• plt.close('all'): closes all plots including those that may not have been displayed
• plt.clf : closes current figure

FORMATTING

Labels

plt.xlabel('x_name_string')
plt.ylabel('y_name_string'')
plt.title('title_string')

Legend

• plt.legend(['label1', 'label2'], loc=#) : loc positions the legend (defult is best), if specify in .plot() dont need to include list of labels
  • 0 : best
  • 1 : upper right
  • 2 : upper left
  • 3 : lower left
  • 4 :lower right
  • 5 : right
  • 6 : center left
  • 7 : center right
  • 8 : lower center
  • 9 : upper center
  • 10 : center

Set axis

plt.axis(arg)

• plt.axis([min_x, max_x, min_y, max_y]) : zoom in on graph with specs
• plt.axis('equal') : shape a circle properly i.e. pie chart

Modify ticks

ax = plt.subplot(args)
plt.plot(args)
ax.set_xticks([values_for_ticks])
ax.set_yticks([values_for_ticks], rotation=#)

• ax : an axes object, allows to modify axes in subplot
• ax.set_xticklabels([label_names_for_ticks])
• ax.set_yticklabels([label_names_for_ticks])

Subplot

plt.subplot(#rows, #cols. index_plot_creating)
plt.plot()

• display two lines side-by-side in different axis, same figure
• any plt.plot() called after subplot will create line in that subplot
• plt.subplots_adjust(op_args=value) customise spacing between subplots
  • left : left-side margin (default of 0.125)
  • right : right-side margin (default of 0.9)
  • bottom : bottom margin, with a default of 0.1
  • top : top margin, with a default of 0.9
  • wspace : the horizontal space between adjacent subplots, with a default of 0.2
  • hspace : the vertical space between adjacent subplots, with a default of 0.2

Figure

• is the picture/object that contails all of the subplots plt.figure(figsize=(width, height)) : creates a new figure plt.savefig('name_of_graph.png') : save as png, cbg, pdf

Figures and Subplots

fig = plt.figure()
fig.add_subplot()

• add subplot to a figure

Seaborn color

sns.set()
for col in 'xy':
  plt.hist(data[col], normed=True, alpha=0.5)

LINE GRAPH

plt.plot(x_values, y_values, color='color_name', linestyle='style', marker='marker', label='for_legend', linewidth=#width)

• visualise how a variable changes over time
• plt.plot(x_values, y_values) : values1 vs. values2 = y_values vs. x_values

Multiple

plt.plot(x1, y1)
plt.plot(x2, y2)

• call plot twice to have multiple line graphs on same axis

Line Format

• color can be HTML color name or HEX name
• linestyle changes the line
  • -- : dashed
  • : : dotted
  • : no line
• marker sets style for data points
  • o : circle
  • s : square
  • * : star

Shade Error

y_lower = [i - error_value for i in y_values]
y_upper = [i + error_value for i in y_values]
plt.fill_between(x_values, y_lower, y_upper, aplha=transp_amount)

• create shaded error region, then plot line over it
• alpha transparency amount between 0 and 1

BAR CHART

plt.bar(x_values, y_values)

• compare multiple categories of data
• the x-values — a list of x-positions for each bar
• the y-values — a list of heights for each bar

Multiple

x_values1 = [t*element + w*n for element
             in range(d)]

• do this for both datasets (changing n for each)
  • n = # This is our first dataset (out of 2)
  • t = # Number of datasets
  • d = # Number of sets of bars
  • w = # Width of each bar (default is 0.8)

Stacked

plt.bar(x1, y1)
plt.bar(x2, y2, bottom=y1)

• first set normal (will be bottom), second set with bottom set to first y (will be top)

Error Bars

plt.bar(x_values, y_values, yerr=error_value, capsize=cap_width)

• caps: horizontal lines at top and bottom
• set error_value to be a list to have a different amount of error for each bar

PIE CHART

plt.pit(data_list, 
        labels=category_list, 
        autopct='format', 
        colors=list_colors
        shadow=True
        startangle=#
        pctdistance=#
        explode=explode)

• display elements as a proportions of a whole
• plt.axis('equal') : almost always need to use to reshape properly
• autopct : shows percentage on each slice with set format
  • ‘%0.2f’ — 2 decimal places
  • ‘%0.2f%%’ — 2 decimal places, but with a percent sign at the end
  • ‘%d%%’ — rounded to the nearest int and with a percent sign at the end
• explode = (0.1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0) : the offeset for each slice i.e. apply 0.1 offset to slice one so it is seperated from the rest of the chart

HISTOGRAM

plt.hist(dataset, range=(min, max), bins=#bins, edgecolor = 'black')

• shows how many values in a dataset fall bbetween sets of numbers
• width of each bin is distance been min and max (same)
• height is the number of elements that fall in that bin
• default #bins = 10

Multiple

plt.hist(a, range=(min, max), bins=#bins, alpha=#transp, normed=true)
plt.hist(b, range=(min, max), bins=#bins, alpha=#transp, normed=true)

• call plt.hist() twice before plt.show()
• alpha=#transp : 0 is entirely transparent, 1 is opaque
• histtype='step : just show outline instead of alpha
• normed=True : normalise histograms (good when different sample sizes)
• linewidth=#width

SCATTER

plt.scatter(x_values, y_values)

3D

fig_3d = plt.figure()
fig_3d.add_subplot(1,1,1,projection='3d')
plt.scatter(x,y,z)

BOXPLOTS

plt.boxplot(dataset)
or
plt.boxplot([dataset1, dataset2])

• one of the most common ways to visualize a dataset

• give you a sense of the central tendency and spread of the data.

• line in the center of the box = median (may not be exactly in middle of box)
• edges of the box = first and third quartiles.
• box length = interquartile range (the middle 50% of data)
• whiskers = spread of dataset. There are many different ways to calculate the length of the whiskers.
  • closest point that is 1.5 * IQR from Q1 and Q3 (eg. Q1 - (1.5*IQR))
  • min and max
  • one standard deviation from mean
• dots beyond whiskers = outliers