Histogram of answers submitted by students for each problem

Import a few dependencies.



In [1]:

    
import requests
from collections import defaultdict

Retrieve the JSON data from the insights module using the query interface.



In [2]:

    
r = requests.get('http://localhost:8002/query/all_problem_answers')

The response contains JSON organized in a flat structure, we will group it by problem later.



In [3]:

    
r.json()[:3]









    Out[3]:





[{u'_id': {u'$oid': u'529500c1701cd9cbbba8708e'},
  u'answer': u'choice_2',
  u'count': 16,
  u'question': u'i4x-edX-E101-problem-a0effb954cca4759994f1ac9e9434bf4_3_1'},
 {u'_id': {u'$oid': u'529500c1701cd9cbbba8708f'},
  u'answer': u'blue',
  u'count': 15,
  u'question': u'i4x-edX-E101-problem-a0effb954cca4759994f1ac9e9434bf4_2_1'},
 {u'_id': {u'$oid': u'529500c1701cd9cbbba87090'},
  u'answer': [u'choice_0', u'choice_2'],
  u'count': 13,
  u'question': u'i4x-edX-E101-problem-a0effb954cca4759994f1ac9e9434bf4_4_1'}]

Group the data by problem.



In [4]:

    
problems = defaultdict(dict)
for bucket in r.json():
    question = bucket['question']
    answer = bucket['answer']
    if isinstance(answer, list):
        answer = ','.join(answer)
    problems[question].update({answer: bucket['count']})

Now the data is grouped by problem with one entry in a dict for each bucket.



In [5]:

    
problems.iteritems().next()









    Out[5]:





(u'i4x-edX-E101-problem-a0effb954cca4759994f1ac9e9434bf4_3_1',
 {u'choice_0': 3, u'choice_1': 5, u'choice_2': 16})

Render a small histogram for every problem on the same axes.



In [7]:

    
fig, ax = plt.subplots()

# Semi-arbitrary
bar_width = 0.25

# The maximum count of any response.  Used to scale the y-axis.
max_value = 0

# Keeps track of the location along the x-axis where the histogram should be
# rendered.  Note this increases with each rendered problem.
problem_x_offset = 0

# Stores the labels for the bars.  Will appear below the x-axis.
x_tick_locations = []
x_tick_labels = []

for problem, values in problems.iteritems():
    # Calculate the left coordinates of each of the bars.
    # problem_x_offset is the left coordinate of the problem
    # The resulting list should look like: 
    # [problem_x_offset, problem_x_offset + bar_width, problem_x_offset + (bar_width*2), ...]
    bar_index = np.arange(0, len(values)*bar_width, bar_width) + problem_x_offset
    
    # Calculate the maximum count seen across all problems, this will
    # be used to determine the scale of the y-axis.
    max_value_for_this_problem = max(values.values())
    max_value = max(max_value_for_this_problem, max_value)
    
    # Generate the plot
    plt.bar(bar_index, values.values(), bar_width,
            label=problem,
            color=np.random.rand(3,))  # Use a random color
    
    # Gather labels for the x-axis with the various responses seen
    x_tick_locations.extend(bar_index + (bar_width/2))
    x_tick_labels.extend(values.keys())
    
    # Put an empty bar_width of space between histograms for each problem
    problem_x_offset = bar_index.max() + (bar_width*2)

# Render the x-axis labels at an angle with the right most letter centered under the bar
plt.xticks(x_tick_locations, x_tick_labels, rotation=50, ha='right')

# Render some empty space above the top of the highest bar
plt.ylim(0, max_value*1.25)

# Increase the size of the plot
fig.set_size_inches(16, 14)

plt.legend()
plt.show()