#!/usr/bin/env python
# coding: utf-8

# # Overview of usage data for the Math Exam Resources Wiki

# <table border="0">
# <tr>
# <th> Main page: </th> <td> <a href="http://wiki.ubc.ca/Science:Math_Education_Resources"> wiki.ubc.ca/Science:Math_Education_Resources </a> </td>
# <td rowspan="4"> <img src="https://raw.github.com/MER-wiki/google-analytics/master/Math_kid.jpg" width=290> </td>
# </tr>
# <tr>
# <th> Contributors portal: </th> <td> <a href="http://wiki.ubc.ca/Science:MER"> wiki.ubc.ca/Science:MER </a>       </td>
# </tr>
# <tr>
# <th> Source of data: </th> <td> google-analytics from CTLT </td>
# </tr>    
# <th> Contact info: </th> <td> mer-wiki (at) math (dot) ubc (dot) ca </td>
# </tr>
# </table>

# #### Technical background
# 
# The python helper function are in their own py file to keep this summary relatively readable.

# In[2]:


get_ipython().run_cell_magic('capture', '', "import numpy as np\nimport matplotlib.pyplot as plt\n%pylab inline\nfrom MERhelpers import *\n\n#Define primary wiki colours\nmainColour = '#2F4F4F'; secondColour1 = '#666666'; secondColour2 = '#7491A3';\n\n#XKCD-ify the plots\nmatplotlib.pyplot.xkcd(scale=0.5, length=100, randomness=5); matplotlib.rcParams['axes.color_cycle'] = [mainColour,secondColour1, secondColour2]; matplotlib.rcParams['font.size'] = 10\n")


# Load data set, then do some clean up to reduce down to only valid course names.

# In[2]:


dict_all,date_clicks,num_clicks = data_to_dict_clickdates_clickscount('Analytics_all.csv')
clean_dict(dict_all,['MATH100','MATH101','MATH102','MATH103','MATH104','MATH105','MATH110','MATH152', 'MATH200','MATH215','MATH220','MATH221','MATH253','MATH257','MATH307','MATH312','MATH437'])
current_term_dict = dict(dict_all)
clean_dict(current_term_dict,['MATH100'])#,'MATH103','MATH105','MATH110','MATH152','MATH215','MATH220'])


# ## Total number of clicks

# Total time series, with the dates of the exams highlighted.

# In[3]:


plt.figure(figsize(16,4)); plot_total_clicks_time_series(date_clicks, num_clicks,False,True)
plt.figure(figsize(16,4)); plot_total_clicks_time_series(date_clicks, num_clicks,True,False)


# Same data (except summer terms) overlayed for comparison.

# In[4]:


Term2_2011 = num_clicks[date_clicks.index('1/1/12'):date_clicks.index('5/1/12')]; Term1_2012 = num_clicks[date_clicks.index('9/1/12'):date_clicks.index('1/1/13')]; Term2_2012 = num_clicks[date_clicks.index('1/1/13'):date_clicks.index('5/1/13')]; Term1_2013 = num_clicks[date_clicks.index('9/1/13'):date_clicks.index('1/1/14')]; Term2_2013 = num_clicks[date_clicks.index('1/1/14'):date_clicks.index('4/30/14')]; Term1_2014 = num_clicks[date_clicks.index('9/1/14'):date_clicks.index('1/1/15')];
Finals_Term2_2011 = num_clicks[date_clicks.index('4/1/12'):date_clicks.index('5/1/12')]; Finals_Term1_2012 = num_clicks[date_clicks.index('12/1/12'):date_clicks.index('1/1/13')]; Finals_Term2_2012 = num_clicks[date_clicks.index('4/1/13'):date_clicks.index('5/1/13')]; Finals_Term1_2013 = num_clicks[date_clicks.index('12/1/13'):date_clicks.index('1/1/14')]; Finals_Term2_2013 = num_clicks[date_clicks.index('4/1/14'):date_clicks.index('4/30/14')]; Finals_Term1_2014 = num_clicks[date_clicks.index('12/1/14'):date_clicks.index('1/1/15')];
Non_Finals_Term2_2011 = sum(Term2_2011)-sum(Finals_Term2_2011); Non_Finals_Term1_2012 = sum(Term1_2012)-sum(Finals_Term1_2012); Non_Finals_Term2_2012 = sum(Term2_2012)-sum(Finals_Term2_2012); Non_Finals_Term1_2013 = sum(Term1_2013)-sum(Finals_Term1_2013); Non_Finals_Term2_2013 = sum(Term2_2013)-sum(Finals_Term2_2013); Non_Finals_Term1_2014 = sum(Term1_2014)-sum(Finals_Term1_2014)


# Total number of clicks, during term and finals

# In[5]:


fig = plt.figure(figsize(8,4))
ax = fig.add_subplot(111)
plt.bar(np.arange(18),[Non_Finals_Term2_2011, sum(Finals_Term2_2011),sum(Term2_2011), Non_Finals_Term1_2012, sum(Finals_Term1_2012), sum(Term1_2012), Non_Finals_Term2_2012, sum(Finals_Term2_2012), sum(Term2_2012), Non_Finals_Term1_2013, sum(Finals_Term1_2013), sum(Term1_2013), Non_Finals_Term2_2013, sum(Finals_Term2_2013), sum(Term2_2013), Non_Finals_Term1_2014, sum(Finals_Term1_2014), sum(Term1_2014)], 
        color=[mainColour,mainColour,mainColour,secondColour2,secondColour2,secondColour2])
xTickMarks = ['Jan-Mar 2012','Apr 2012', 'Term (11-2)','Sep-Nov 2012', 'Dec 2012','Term (12-1)','Jan-Mar 2013','Apr 2013','Term (12-2)','Sep-Nov 2013', 'Dec 2013','Term (13-1)','Jan-Mar 2014', 'Apr 2014', 'Term (13-2)', 'Sep-Nov 2014', 'Dec 2014', 'Term (14-1)']
ind = np.arange(18); ax.set_xticks(ind+0.35); xtickNames = ax.set_xticklabels(xTickMarks)
plt.ylabel('(thousands)', size='x-large'); plt.setp(xtickNames, rotation=80, fontsize=14); plt.title('Total number of clicks during term and finals',size=14); plt.gca().set_yticklabels(['0','50','100','150','200','250', '300', '350', '400', '450'],size='x-large')

handles, labels = ax.get_legend_handles_labels(); plt.savefig('total_number_of_clicks_term_finals_bar.png', bbox_inches='tight'); plt.show()


# ## Where do users spend their time?

# ### Current Term

# In[19]:


dict_current, date_clicks_current, num_clicks_current = data_to_dict_clickdates_clickscount('Analytics_Winter2015.csv')
clean_dict(dict_current, ['MATH100','MATH101','MATH102','MATH103','MATH104','MATH105','MATH110','MATH152', 'MATH200','MATH215','MATH220','MATH221','MATH253','MATH257','MATH307','MATH312','MATH437'])
matplotlib.rcParams['font.size'] = 16

plt.figure(figsize(8,8))
plot_pie_chart_time_spent_on_course(dict_current,0.02,[mainColour,secondColour1,secondColour2])
plt.savefig('total_time_per_course.png', bbox_inches='tight')  
plt.show()


# ### Overall

# In[7]:


matplotlib.rcParams['font.size'] = 16

plt.figure(figsize(8,8))
plot_pie_chart_time_spent_on_course(dict_all,0.02,[mainColour,secondColour1,secondColour2])
plt.savefig('total_time_per_course.png', bbox_inches='tight')  
plt.show()


# ### Clicks per exam

# How many previous exams do students use to prepare? We would expect the most recent exam to have the most visits. However, some exam were uploaded later.

# In[12]:


plt.figure(figsize(15,50))
plt.subplots_adjust(hspace=0.6)
for count, course in enumerate(list_courses(dict_current)):
    if not course in ['MATH101']:
        continue
    plt.subplot(3,1,count+1)
    plot_visit_duration_dist(dict_current, course, 'bar')
plt.subplots_adjust(left = 0.125, right= 0.9, bottom = 0.5, top = 0.9, wspace = 0.5, hspace = 0.6)


# ### Time to solution view
# 
# Using analytics provided by the UBC wiki webmasters, MER has information about the difference in time between when students load a page and click to open the question box. A cursory glance suggests two distinct groups of clicks. The first group of clicks occur at times less than 25 secs and the second group appears to have a mode of approximately 2 mins. The data is from the time period from 04/05/2014 (MM/DD/YYYY) to 05/05/2014, encompasssing the April 2014 Exam Period.

# In[9]:


del_t = load_delta_t('raw_data/Analytics wiki.ubc.ca Top Events 20140405-20140505.csv'); del_t.extend(load_delta_t('raw_data/Analytics wiki.ubc.ca Top Events 20140405-20140505-2.csv')); del_t.extend(load_delta_t('raw_data/Analytics wiki.ubc.ca Top Events 20140405-20140505-3.csv')); del_t.extend(load_delta_t('raw_data/Analytics wiki.ubc.ca Top Events 20140405-20140505-4.csv')); del_t.extend(load_delta_t('raw_data/Analytics wiki.ubc.ca Top Events 20140405-20140505-5.csv')); del_t = np.asarray(del_t)
print 'Total counted delta_t events:', len(del_t)


# In[10]:


plt.figure(figsize(10,4))
plot_data = del_t[del_t>0]/1000.0 # Three data points are negative, funny. But so few that we don't need to worry about it.
num_bins = 60; plt.hist(np.log10(plot_data), num_bins, facecolor=secondColour2)
plt.xticks(np.log10([1,5,10,25,2*60,4*60,10*60,30*60, 2*60*60, 5*60*60]),['0','5s','10s', '25s', '2m', '4m', '10m', '30m', '2h', '5h']); plt.xlim(0,np.log10([5*60*60]))
plt.title('Time between loading question page and opening solution box'); plt.ylabel('frequency.'); plt.xlabel('time (logarithmic)'); plt.tight_layout(); plt.savefig('2013_Term2_delta_t.png',bbox_inches='tight'); plt.show()


# [Correlation analysis of exam score vs. easiness rating](http://nbviewer.ipython.org/github/BernhardKonrad/correlation_analysis/blob/master/MER_correlation_analysis.ipynb)