Faculty base in few Indian Institutes
Lets' visualize
Occasionally, I get the urge to know more about the Indian Institutes of National Importance. This time, the urge was fuelled by seeing the NIRF 2021 datasheet prepared and hosted by some of the institutes on their website. Staring at the datasheets, I thought it's better to visualize it. You may find this useful for
- understanding the institutes on dimensions shown in the image below
- evaluating the institutes if you are a faculty aspirant
- cheering up and/or bringing a change
Note that this presentation has my own bias of visualizing only along certain dimensions. I encourage you to go deeper if you are interested in something I don't discuss below. At the end you will find links to the data sheets I have used for this presentation.
Step 1: I collected the NIRF 2021 data hosted by a few institutes of our interest. Note that this is 2021 data, and as I understand, this has been submitted to the Ministry of Education and hence, we can consider them authentic. The NIRF 2021 rankings are yet to be released. Thanks to the meticulous data preparation by a few institutes, I could easily download this data from the institute websites. We will focus on 13 institutes. These are:
- Indian Institute of Science (iisc)
- IIT Bombay (iit_bom), IIT Kharagpur (iit_kgp), IIT Madras (iit_mad), IIT Kanpur (iit_kan), IIT Delhi (iit_delhi), IIT Varanasi (iit_bhu), IIT Roorkee (iit_roor)
- IIT Jodhpur (iit_jodhpur), IIT Indore (iit_ind), IIT Bhubaneswar (iit_bbs), IIT Mandi (iit_mandi), IIT Hyderabad (iit_hyd)
Below you can see these institute locations overlaid on a map. The idea of IIT was conceived around 1947 to enable India's technology focussed societal progress after the second world war. It was decided to create one of these in the north, south, east and west of India. However, now we have them spread across many more places, with most of the states having one IIT (not all are shown in the map below). This is indeed very good as it helps to cater to the large population of India.
#collapse
import PyPDF2
import glob
import numpy as np
import pandas as pd
from geopy.geocoders import Nominatim
import matplotlib.pyplot as plt
plt.rcParams.update({'font.size': 12})
from mpl_toolkits.basemap import Basemap
import seaborn as sns
sns.set() # Use seaborn's default style to make attractive graphs
sns.set_style("white")
sns.set_style("ticks")
institutes = ['iisc','iit_kgp','iit_bhu','iit_delhi','iit_mad','iit_bom','iit_kan','iit_roor',
'iit_ind','iit_hyd','iit_mandi','iit_jodh','iit_bbs']
# make map
geolocator = Nominatim(user_agent="geoapiExercises")
def geolocate(country):
try:
# Geolocate the center of the country
loc = geolocator.geocode(country)
# And return latitude and longitude
return (loc.latitude, loc.longitude)
except:
# Return missing value
return np.nan
city_names = ['bangalore', 'kharagpur','varanasi','delhi','chennai','mumbai','kanpur','roorkee',
'indore','hyderabad','mandi','jodhpur','bhubaneswar']
df = {}
df['city_name'] = city_names
df = pd.DataFrame.from_dict(df)
df['lat'] = 0
df['long'] = 0
for i in range(len(df)):
temp = geolocate(df['city_name'][i])
df.loc[i,'lat'] = temp[0]
df.loc[i,'long'] = temp[1]
# Make the india map and overlay locations
fig = plt.subplots(figsize=[12,12])
# m = Basemap(llcrnrlon=-180, llcrnrlat=-65, urcrnrlon=180, urcrnrlat=80)
m = Basemap(llcrnrlon=60, llcrnrlat=5, urcrnrlon=100, urcrnrlat=37)
m.drawmapboundary(fill_color='#FFF', linewidth=0)
m.fillcontinents(color='grey', alpha=0.25)
m.drawcoastlines(linewidth=0.1, color="white")
# prepare a color for each point depending on the continent.
df['labels_enc'] = pd.factorize(df['city_name'])[0]
# Add a point per position
m.scatter(x=df['long'], y=df['lat'], s=50,alpha=1, c=df['labels_enc']/100, cmap="Set1")
for i in range(len(df)):
x = df.loc[i,'long']+0.5
y = df.loc[i,'lat']
plt.text(x, y, institutes[i])
plt.show()
Step 2: I did a little bit of coding (in python) exercise to extract data from the PDF documents collected in Step 1. This data extraction step is critical as it helps to load the data into a plotting tool!
Step 3: Rest is just plotting (again, using python).
Note: I will be analyzing data of faculty at three levels, namely, Assistant, Associate, and Full Professor. Also, I have skipped the Institute Director (1 nos.) from the analysis.
So, what do we see!
#collapse
# path to PDFs
path_nirf = '/Users/neeks/Desktop/Documents/work/code/python_codes/notebookCodes/institutes/data/nirf_pdfs/'
# some function defs
def get_curated_rows(pdfReader):
# get faculty listing start page
for i in range(pdfReader.numPages):
pageObj = pdfReader.getPage(i)
lines = pageObj.extractText()
if 'Faculty Details' in lines:
page_indx = i
break
all_rows = []
for i in range(page_indx,pdfReader.numPages):
pageObj = pdfReader.getPage(i)
if i == page_indx:
lines = pageObj.extractText().split('Association type')[1]
else:
lines = pageObj.extractText()
line_breaks = ['Regular', 'Visiting', 'Other']
temp = lines
for text in line_breaks:
temp = temp.replace(text,';')
all_rows.extend(temp.split(';'))
# remove contractual rows
curated_rows = []
for text in all_rows:
if len(text) == 0:
continue
elif 'Adhoc' in text:
temp = text.split('Adhoc /Contractual')
for j in range(len(temp)):
temp_1 = temp[j]
if 'Professor' in temp_1:
curated_rows.append(temp_1)
else:
if 'Professor' in text:
curated_rows.append(text)
return curated_rows
def get_parsed_dict(curated_rows, code):
# make dictionary
keys = ['name','age','designation','gender','degree','months','joining_day',
'joining_month','joining_year','left','institute']
data_dict = {}
for key in keys:
data_dict[key] = []
for text in curated_rows:
# remove serial number
for i in range(len(text)):
if (i == 0) & (text[i]==' '):
continue
if not(text[i].isdigit()):
break
text = text[i:]
# search name
for i in range(len(text)):
if text[i].isdigit():
break
data_dict['name'].append(text[:i])
text = text[i:]
# search age
for i in range(len(text)):
if not(text[i].isdigit()):
break
data_dict['age'].append(int(text[:i]))
text = text[i:]
# search designation
text = text.split('Professor')
if len(text[0]) == 0:
data_dict['designation'].append('Professor')
else:
data_dict['designation'].append(text[0])
text = text[1]
# search for gender, degree
if 'Ph.D' in text:
data_dict['gender'].append(text.split('Ph.D')[0])
data_dict['degree'].append('Ph.D')
text = text.split('Ph.D')[1]
elif 'M.Tech' in text:
data_dict['gender'].append(text.split('M.Tech')[0])
data_dict['degree'].append('M.Tech')
text = text.split('M.Tech')[1]
elif 'B.Tech' in text:
data_dict['gender'].append(text.split('B.Tech')[0])
data_dict['degree'].append('B.Tech')
text = text.split('B.Tech')[1]
elif 'Master of Design' in text:
data_dict['gender'].append(text.split('Master of Design')[0])
data_dict['degree'].append('M.Des')
text = text.split('Master of Design')[1]
elif 'PGID' in text:
data_dict['gender'].append(text.split('PGID')[0])
data_dict['degree'].append('PGID')
text = text.split('PGID')[1]
elif 'PGDBM' in text:
data_dict['gender'].append(text.split('PGDBM')[0])
data_dict['degree'].append('PGDBM')
text = text.split('PGDBM')[1]
elif 'M.Sc.(Engg)' in text:
data_dict['gender'].append(text.split('M.Sc.(Engg)')[0])
data_dict['degree'].append('M.Sc.(Engg)')
text = text.split('M.Sc.(Engg)')[1]
elif 'M.Sc.' in text:
data_dict['gender'].append(text.split('M.Sc.')[0])
data_dict['degree'].append('M.Sc.')
text = text.split('M.Sc.')[1]
elif 'MFA(Fine Arts)' in text:
data_dict['gender'].append(text.split('MFA(Fine Arts)')[0])
data_dict['degree'].append('MFA(Fine Arts)')
text = text.split('MFA(Fine Arts)')[1]
elif 'P.G.Diploma' in text:
data_dict['gender'].append(text.split('P.G.Diploma')[0])
data_dict['degree'].append('P.G.Diploma')
text = text.split('P.G.Diploma')[1]
elif 'M. Phil' in text:
data_dict['gender'].append(text.split('M. Phil')[0])
data_dict['degree'].append('M. Phil')
text = text.split('M. Phil')[1]
elif 'M.Arch.' in text:
data_dict['gender'].append(text.split('M.Arch.')[0])
data_dict['degree'].append('M.Arch.')
text = text.split('M.Arch.')[1]
elif 'MBA' in text:
data_dict['gender'].append(text.split('MBA')[0])
data_dict['degree'].append('MBA')
text = text.split('MBA')[1]
elif 'M.A' in text:
data_dict['gender'].append(text.split('M.A')[0])
data_dict['degree'].append('M.A')
text = text.split('M.A')[1]
elif 'M.S' in text:
data_dict['gender'].append(text.split('M.S')[0])
data_dict['degree'].append('M.S')
text = text.split('M.S')[1]
elif 'M.E.' in text:
data_dict['gender'].append(text.split('M.E.')[0])
data_dict['degree'].append('M.E.')
text = text.split('M.E.')[1]
elif 'B.E' in text:
data_dict['gender'].append(text.split('B.E')[0])
data_dict['degree'].append('B.E')
text = text.split('B.E')[1]
# search months
for i in range(len(text)):
if (i == 0) & (text[i]==' '):
continue
if not(text[i].isdigit()):
break
data_dict['months'].append(int(text[:i]))
text = text[i:]
# joining date
for i in range(len(text)):
if text[i].isdigit():
break
text = text[i:]
data_dict['joining_day'].append(int(text.split('-')[0]))
data_dict['joining_month'].append(int(text.split('-')[1]))
data_dict['joining_year'].append(int(text.split('-')[2][:4]))
# left or continues
text = text.split('-')[2][4:]
if len(text)>0:
if text[0].isdigit():
data_dict['left'].append(1)
else:
data_dict['left'].append(0)
# institute code
data_dict['institute'].append(code)
return data_dict
def get_admin_info(pdfReader, code):
admin_data = {}
keys = ['PhDs_full','PhDs_part', 'staff_salaries', 'institute']
for key in keys:
admin_data[key] = []
# search for PhD details page
for i in range(pdfReader.numPages):
pageObj = pdfReader.getPage(i)
lines = pageObj.extractText()
if 'Ph.D Student Details' in lines:
page_indx = i
break
pageObj = pdfReader.getPage(i)
lines = pageObj.extractText()
text = lines.split("Total StudentsFull Time")[1]
# get PhD Full time count
for i in range(len(text)):
if (i == 0) & (text[i]==' '):
continue
if not(text[i].isdigit()):
break
admin_data['PhDs_full'].append(int(text[:i]))
text = text[i:].split('Part Time')[1]
# get PhD Part time count
for i in range(len(text)):
if (i == 0) & (text[i]==' '):
continue
if not(text[i].isdigit()):
break
admin_data['PhDs_part'].append(int(text[:i]))
# search for salaries page
for i in range(pdfReader.numPages):
pageObj = pdfReader.getPage(i)
lines = pageObj.extractText()
if 'Salaries (Teaching and Non Teaching staff)' in lines:
page_indx = i
break
pageObj = pdfReader.getPage(i)
lines = pageObj.extractText()
text = lines.split("Salaries (Teaching and Non Teaching staff)")[1]
# get PhD Full time count
for i in range(len(text)):
if (i == 0) & (text[i]==' '):
continue
if not(text[i].isdigit()):
break
admin_data['staff_salaries'].append(int(text[:i]))
admin_data['institute'].append(code)
return admin_data
def remove_items(test_list, item):
# using list comprehension to perform the task
res = [i for i in test_list if i != item]
return res
def name_filter(name):
temp = name.split()
if len(temp[0])<3:
temp_1 = temp[1]
else:
if (temp[0] == 'SMT') | (temp[0] == 'Prof'):
temp_1 = temp[1]
else:
temp_1 = temp[0]
return temp_1
# main code starts here
institutes = ['iisc','iit_kgp','iit_bhu','iit_delhi','iit_mad','iit_bom','iit_kan','iit_roor',
'iit_ind','iit_hyd','iit_mandi','iit_jodh','iit_bbs']
cnt = 0
for insitute in institutes:
file_name = glob.glob(path_nirf+insitute+'*')[0]
pdfFileObj = open(file_name, 'rb')
pdfReader = PyPDF2.PdfFileReader(pdfFileObj)
# get admmin df
data = get_admin_info(pdfReader, insitute)
df_admin_temp = pd.DataFrame.from_dict(data)
if cnt == 0:
df_admin = df_admin_temp.copy()
else:
df_admin = pd.concat([df_admin, df_admin_temp])
# get faculty df
rows = get_curated_rows(pdfReader)
data = get_parsed_dict(rows, insitute)
df_fac_temp = pd.DataFrame.from_dict(data)
if cnt == 0:
df_fac = df_fac_temp.copy()
else:
df_fac = pd.concat([df_fac, df_fac_temp])
cnt = cnt+1
#collapse
# plot PhD count
data = {}
keys_1 = ['PhDs_full','PhDs_part']
with plt.xkcd():
fig = plt.subplots(figsize=[16,8])
ax = plt.subplot(1,1,1)
FS = 14
clr = ['cornflowerblue','lightcoral']
for i in range(len(keys_1)):
if i == 0:
ax.bar(np.arange(len(institutes))+0.1*i,df_admin['PhDs_full'],width=0.3,color=clr[i],
label=keys_1[i]+str('-time'))
else:
ax.bar(np.arange(len(institutes))+0.1*i,df_admin['PhDs_part'],width=0.3,color=clr[i],
label=keys_1[i]+str('-time'))
plt.xticks(np.arange(0,len(institutes))+.1,institutes,rotation=0, fontsize=FS)
ax.grid(color='gray', linestyle='--', linewidth=1,alpha=.3)
plt.legend(frameon=False,fontsize=FS)
plt.xticks(fontsize=FS)
plt.yticks(fontsize=FS)
ax.set_ylabel('SCHOLAR COUNT \n (as of 2020-21)', fontsize=FS)
ax.spines['right'].set_visible(False)
ax.spines['top'].set_visible(False)
# fmt = '.pdf'
# if fig_save:
# ax.figure.savefig(path_store_figure+"dicova_track_2_dur"+fmt, bbox_inches='tight')
plt.show()
- I used to think that IITs cater mostly to undergraduate education. After seeing the plot, my assumption was found wrong. Some of the IITs have close to (or more than) 2500 PhD Scholars (in 2020). This is huge, and even more than that of IISc, a leading research institute of India. With a mix of good undergraduate and PhD Scholar population in these IITs, these institutes are in a unique position to offer a dynamic and happening environment for research and education.
- IIT-BHU is falling behind in PhD Scholars count. I don't know why. The new IITs (ii_hyd, iit_ind, iit_jodh, iit_bbs, iit_mandi) were established in 2008-09, and exhibit a good count of PhD Scholars. Hopefully, this will increase. iit_hyd is scaling up really well
- Part-time PhD Scholars are mainly in institutes located in metros (except, iit_roor). Is it because of easy accessibility?
#collapse
# plot all salary exoense
data = {}
keys_1 = ['staff_salaries']
val_fac_count = []
for institute in institutes:
val_fac_count.append(len(df_fac[(df_fac['institute']==institute) & (df_fac['left']==0)]))
with plt.xkcd():
fig = plt.subplots(figsize=[16,7])
ax = plt.subplot(1,1,1)
FS = 14
clr = ['cornflowerblue']
ax.bar(np.arange(len(institutes))+0.25,val_fac_count,width=0.3,color=clr[0])
ax.grid(color='gray', linestyle='--', linewidth=1,alpha=.3)
ax.set_ylabel('FACULTY COUNT (Assis.+Assoc.+Full Prof.)\n (as of Dec. 2020)')
plt.xticks(np.arange(0,len(institutes))+.1,institutes,rotation=0, fontsize=FS)
plt.xticks(fontsize=FS)
plt.yticks(fontsize=FS)
ax.spines['top'].set_visible(False)
ax.spines["right"].set_visible(False)
# fmt = '.pdf'
# if fig_save:
# ax.figure.savefig(path_store_figure+"dicova_track_2_dur"+fmt, bbox_inches='tight')
plt.show()
The well established institutes have more than 400 faculty (just to remind, I am counting Asst.+Assoc.+Prof. only).
- Not all old IITs have same count of faculty. The iit_kgp and iit_bom shoot close to 700! Why is iit_kan lagging in this?
- Although IISc is lower on undergraduate student count (not shown here), the faculty strength is (relatively) quite good, at close to 450. Further, it is close to that of iit_kan and iit_roor!
- Again, amongst the new IITs, iit_hyd is nicely standing tall.
#collapse
# plot all salary exoense
data = {}
keys_1 = ['staff_salaries']
val_fac_count = []
for institute in institutes:
val_fac_count.append(len(df_fac[(df_fac['institute']==institute)& (df_fac['left']==0)]))
with plt.xkcd():
fig = plt.subplots(figsize=[10,10])
ax = plt.subplot(1,1,1)
FS = 14
clr = ['cornflowerblue','lightcoral']
for i in range(len(institutes)):
ax.scatter(df_admin['staff_salaries'].values[i]/1e7,val_fac_count[i],s=100,alpha=.75)
plt.text(df_admin['staff_salaries'].values[i]/1e7+np.random.randint(-5,5),val_fac_count[i]+np.random.randint(0,10),
institutes[i],fontsize=FS-2, rotation=np.random.randint(-90,90))
ax.plot([0,700],[0,700],'--',c='black',alpha=0.5)
plt.xticks(fontsize=FS)
plt.yticks(fontsize=FS)
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)
ax.set_xlabel('SALARY EXPENSE (teaching and non-teaching) [in Cr. INR]')
ax.set_ylabel('FACULTY COUNT (Assis.+Assoc.+Full Prof.)')
ax.grid(color='gray', linestyle='--', linewidth=1,alpha=.3)
ax.set_xlim([0,750])
ax.set_ylim([0,750])
# fmt = '.pdf'
# if fig_save:
# ax.figure.savefig(path_store_figure+"dicova_track_2_dur"+fmt, bbox_inches='tight')
plt.show()
- Is iit_kgp being underpaid? Franknly, this is puzzling.
- or, is it that iit_mad and iisc have a huge staff count outside the faculty pool of Asst.+ Assoc.+Full Prof. and hence, huge salary expense.
#collapse
# exclude faculty who have left
df = df_fac[df_fac['left']==0].copy()
# plot age (without gender)
data = {}
keys = ['age','gender','designation']
for key in keys:
data[key] = []
for institute in institutes:
for key in keys:
data[key].append(df[(df['institute']==institute)][key].values)
with plt.xkcd():
fig = plt.subplots(figsize=[16,7])
ax = plt.subplot(1,1,1)
FS = 14
sns.boxplot(data = data['age'], whis = np.inf,width = 0.2,color=(.3,.8,1),palette="muted")
sns.swarmplot(data = data['age'], color='gray',alpha=0.4,palette="muted")
# sns.violinplot(data = data['age'], palette="muted")
plt.xticks(np.arange(0,len(institutes)),institutes,rotation=0, fontsize=FS-2)
ax.grid(color='gray', linestyle='--', linewidth=1,alpha=.3)
plt.xticks(fontsize=FS)
plt.yticks(fontsize=FS)
ax.set_ylabel('AGE [in yrs]', fontsize=FS)
ax.spines['right'].set_visible(False)
ax.spines['top'].set_visible(False)
# fmt = '.pdf'
# if fig_save:
# ax.figure.savefig(path_store_figure+"dicova_track_2_dur"+fmt, bbox_inches='tight')
plt.show()
Some observations:
- Good spread, from 30s to 65s. In case you are not aware, the retirement age is 65.
- As expected, new IITs have a relatively younger faculty population.
- Why do iit_jodh and iit_bbs have outliers going beyond 65? We will come to this later.
#collapse
# plot age (without gender)
data = {}
keys = ['age','gender','designation']
for key in keys:
data[key] = []
for institute in institutes:
for key in keys:
data[key].append(df[(df['institute']==institute)&(df['designation']=='Assistant')][key].values)
with plt.xkcd():
fig = plt.subplots(figsize=[16,7])
ax = plt.subplot(1,1,1)
FS = 14
sns.boxplot(data = data['age'], whis = np.inf,width = 0.2,color=(.3,.8,1),palette="muted")
sns.swarmplot(data = data['age'], color='gray',alpha=0.4,palette="muted")
# sns.violinplot(data = data['age'], palette="muted")
plt.xticks(np.arange(0,len(institutes)),institutes,rotation=0, fontsize=FS-2)
ax.grid(color='gray', linestyle='--', linewidth=1,alpha=.3)
plt.xticks(fontsize=FS)
plt.yticks(fontsize=FS)
ax.set_ylabel('AGE [in yrs]', fontsize=FS)
ax.spines['right'].set_visible(False)
ax.spines['top'].set_visible(False)
# fmt = '.pdf'
# if fig_save:
# ax.figure.savefig(path_store_figure+"dicova_track_2_dur"+fmt, bbox_inches='tight')
plt.show()
#collapse
# plot age (without gender)
data = {}
keys = ['age','gender','designation']
for key in keys:
data[key] = []
for institute in institutes:
for key in keys:
data[key].append(df[(df['institute']==institute)&(df['designation']=='Associate')][key].values)
with plt.xkcd():
fig = plt.subplots(figsize=[16,7])
ax = plt.subplot(1,1,1)
FS = 14
sns.boxplot(data = data['age'], whis = np.inf,width = 0.2,color=(.3,.8,1),palette="muted")
sns.swarmplot(data = data['age'], color='gray',alpha=0.4,palette="muted")
# sns.violinplot(data = data['age'], palette="muted")
plt.xticks(np.arange(0,len(institutes)),institutes,rotation=0, fontsize=FS-2)
ax.grid(color='gray', linestyle='--', linewidth=1,alpha=.3)
plt.xticks(fontsize=FS)
plt.yticks(fontsize=FS)
ax.set_ylabel('AGE [in yrs]', fontsize=FS)
ax.spines['right'].set_visible(False)
ax.spines['top'].set_visible(False)
# fmt = '.pdf'
# if fig_save:
# ax.figure.savefig(path_store_figure+"dicova_track_2_dur"+fmt, bbox_inches='tight')
plt.show()
#collapse
# plot age (without gender)
data = {}
keys = ['age','gender','designation']
for key in keys:
data[key] = []
for institute in institutes:
for key in keys:
data[key].append(df[(df['institute']==institute)&(df['designation']=='Professor')][key].values)
with plt.xkcd():
fig = plt.subplots(figsize=[16,7])
ax = plt.subplot(1,1,1)
FS = 14
sns.boxplot(data = data['age'], whis = np.inf,width = 0.2,color=(.3,.8,1),palette="muted")
sns.swarmplot(data = data['age'], color='gray',alpha=0.4,palette="muted")
# sns.violinplot(data = data['age'], palette="muted")
plt.xticks(np.arange(0,len(institutes)),institutes,rotation=0, fontsize=FS-2)
ax.grid(color='gray', linestyle='--', linewidth=1,alpha=.3)
plt.xticks(fontsize=FS)
plt.yticks(fontsize=FS)
ax.set_ylabel('AGE [in yrs]', fontsize=FS)
ax.spines['right'].set_visible(False)
ax.spines['top'].set_visible(False)
# fmt = '.pdf'
# if fig_save:
# ax.figure.savefig(path_store_figure+"dicova_track_2_dur"+fmt, bbox_inches='tight')
plt.show()
Some observations:
- iit_mandi is yet to feature a Full Professor. I have excluded the Dean here. The quality of education and research the institute has attained even without this is impressive!
- iit_jodh and iit_bbs have relatively more senior Full Professors. Some go beyond 65. In the datasheet these are categorised as Visiting.
- iit_ind and iit_hyd feature a group of young Full Professors.
#collapse
# plot deisgnation (with gender)
data = {}
keys_1 = ['Male','Female']
keys_2 = ['Assistant','Associate','Professor']
for key_1 in keys_1:
data[key_1] = {}
for key_2 in keys_2:
data[key_1][key_2] = []
for institute in institutes:
for key_1 in keys_1:
for key_2 in keys_2:
data[key_1][key_2].append(len(df[(df['institute']==institute) &
(df['gender']==key_1) & (df['designation']==key_2)]))
with plt.xkcd():
fig = plt.subplots(figsize=[16,7])
ax = plt.subplot(1,1,1)
FS = 14
clr = ['cornflowerblue','tab:red','tab:cyan']
for i in range(len(keys_1)):
for j in range(len(keys_2)):
if i == 0:
ax.bar(np.arange(len(institutes))+0.1*i+0.2*j,data[keys_1[i]][keys_2[j]],width=0.1,color=clr[j],
label=keys_2[j]+str(' (M)'))
else:
ax.bar(np.arange(len(institutes))+0.1*i+0.2*j,data[keys_1[i]][keys_2[j]],width=0.1,color=clr[j],
hatch='//',label=keys_2[j]+str(' (F)'))
plt.xticks(np.arange(0,len(institutes))+.25,institutes,rotation=0, fontsize=FS)
ax.grid(color='gray', linestyle='--', linewidth=1,alpha=.3)
plt.legend(frameon=False,fontsize=FS)
plt.xticks(fontsize=FS)
plt.yticks(fontsize=FS)
ax.set_ylabel('COUNT', fontsize=FS)
ax.spines['right'].set_visible(False)
ax.spines['top'].set_visible(False)
# fmt = '.pdf'
# if fig_save:
# ax.figure.savefig(path_store_figure+"dicova_track_2_dur"+fmt, bbox_inches='tight')
plt.show()
- For a well established institute, (Asst. nos.) < (Assoc. nos.) < (Full Prof. nos) will likely indicate the institute is recruiting, promoting, and retiring in a timely manner. Example, iit_mad!
- Every institute may have their own strategy for recruitment and promotion. Some institutes recruit in bursts or promote based on lots of checks and bounds. These things might impact the unequal count in faculty across designations.
- New IITs, as these are being established, will feature a higher Asst. nos. But wait, why are iit_bbs and iit_jodh Assoc. nos count relatively low? We will see below that for iit_jodh at least, the reason is the increased recent intake.
#collapse
# plot deisgnation (with gender)
data = {}
keys_1 = ['Male','Female']
for key_1 in keys_1:
data[key_1] = []
for institute in institutes:
for key_1 in keys_1:
data[key_1].append(len(df[(df['institute']==institute) &
(df['gender']==key_1)]))
with plt.xkcd():
fig = plt.subplots(figsize=[16,7])
ax = plt.subplot(1,1,1)
FS = 14
clr = ['cornflowerblue','tab:red','tab:cyan']
ax.bar(np.arange(len(institutes)),np.array(data[keys_1[0]])/np.array(data[keys_1[1]]),width=0.4,color=clr[0],
hatch='//'+str(' (F)'))
plt.xticks(np.arange(0,len(institutes)),institutes,rotation=0, fontsize=FS)
ax.grid(color='gray', linestyle='--', linewidth=1,alpha=.3)
plt.xticks(fontsize=FS)
plt.yticks(fontsize=FS)
ax.set_ylabel('MALE-to-FEMALE FACULTY RATIO', fontsize=FS)
ax.spines['right'].set_visible(False)
ax.spines['top'].set_visible(False)
# fmt = '.pdf'
# if fig_save:
# ax.figure.savefig(path_store_figure+"dicova_track_2_dur"+fmt, bbox_inches='tight')
plt.show()
- It is nice to see iit_mandi, and iit_jodh (the two new IITs) doing relatively better.
- New IITs, iit_hyd, iit_ind, iit_bbs, are following the old ones.
- IISc and iit_kan are the poorest in this, and iit_delhi is much better when compared to other old IITs.
A ratio > 5x is not what an institute of national importance should feature. But sadly, we see this here. Is this due to lacking
- quality applicants
- intent in maintaining diversity in recruitment
- encouragement to choose a faculty career at these institutes
- others There is ann urgent need to go deeper into analyzing each of these aspects, discuss and address them with dedication and sincerity.
#collapse
# year of recruitment
# dont exclude faculty who have left
df = df_fac.copy()
x_grid = np.arange(min(df['joining_year']),2021)
y_grid = np.arange(0,len(institutes))
X, Y = np.meshgrid(x_grid, y_grid)
Z_m = np.zeros(X.shape)
Z_f = np.zeros(X.shape)
for i in range(len(institutes)):
for j in range(len(x_grid)):
Z_m[i,j] = len(df[(df['institute']==institutes[i]) &
((df['joining_year']==x_grid[j])) & ((df['gender']=='Male'))])
Z_f[i,j] = len(df[(df['institute']==institutes[i]) &
((df['joining_year']==x_grid[j])) & ((df['gender']=='Female'))])
with plt.xkcd():
fig = plt.figure(figsize=[16,10])
FS = 14
ax = plt.subplot(1,1,1)
clr = 'yellowgreen'
ax.bar(x_grid,np.sum(Z_m[:,:],axis=0),alpha=1,width=1.0,
color=clr, edgecolor=clr, label='MALE')
clr = 'tab:red'
ax.bar(x_grid,np.sum(Z_f[:,:],axis=0),bottom = np.sum(Z_m[:,:],axis=0), alpha=1,width=1.0,
color=clr, edgecolor=clr, label='FEMALE')
ax.grid(color='gray', linestyle='--', linewidth=1,alpha=.3)
plt.ylabel('FACULTY COUNT\n (new joinee only)',fontsize=FS)
plt.xlabel('YEAR',fontsize=FS)
plt.xticks(fontsize=FS)
plt.yticks(fontsize=FS)
ax.spines['right'].set_visible(False)
ax.spines['top'].set_visible(False)
plt.legend(frameon=False,loc='upper left', fontsize=FS)
plt.show()
- Something happened in 1995, and there was a gradual increase in intake
- It got reset in 1999, and again began to gradually rise till 2004.
- In 2008, with starting of new IITs, there was a rise, with signs of saturation in 2016-2017
- It picked up again in 2018.
#collapse
# year of recruitment
x_grid = np.arange(min(df['joining_year']),2021)
y_grid = np.arange(0,len(institutes))
X, Y = np.meshgrid(x_grid, y_grid)
Z_m = np.zeros(X.shape)
Z_f = np.zeros(X.shape)
for i in range(len(institutes)):
for j in range(len(x_grid)):
Z_m[i,j] = len(df[(df['institute']==institutes[i]) &
((df['joining_year']==x_grid[j])) & ((df['gender']=='Male'))])
Z_f[i,j] = len(df[(df['institute']==institutes[i]) &
((df['joining_year']==x_grid[j])) & ((df['gender']=='Female'))])
clr = ['tab:brown', 'tab:pink','tab:gray','tab:red','tab:cyan',
'darkorange','wheat','yellowgreen','cornflowerblue','slateblue',
'lightcoral','aquamarine','tab:purple']
with plt.xkcd():
fig = plt.figure(figsize=[16,12])
FS = 14
ax = plt.subplot(1,1,1)
for i in range(0,Z_m.shape[0]):
ax.bar(x_grid,Z_m[i,:]+Z_f[i,:],bottom=np.sum(Z_m[:i,:]+Z_f[:i,:],axis=0),alpha=1,width=1.0,
color=clr[i], edgecolor=clr[i], label=institutes[i])
ax.grid(color='gray', linestyle='--', linewidth=1,alpha=.3)
plt.ylabel('FACULTY COUNT\n (new joinee only)',fontsize=FS)
plt.xlabel('YEAR',fontsize=FS)
plt.xticks(fontsize=FS)
plt.yticks(fontsize=FS)
ax.spines['right'].set_visible(False)
ax.spines['top'].set_visible(False)
plt.legend(frameon=False,loc='upper left', fontsize=FS)
plt.show()
- Some institutes are always recruiting in good numbers (see iit_mad, and iit_kgp) and some in small numbers (see iisc).
- iit_jodh started slow, and recently, has been recruiting good numbers
#collapse
# year of recruitment
x_grid = np.arange(min(df['joining_year']),2021)
y_grid = np.arange(0,len(institutes))
X, Y = np.meshgrid(x_grid, y_grid)
Z_m = np.zeros(X.shape)
Z_f = np.zeros(X.shape)
for i in range(len(institutes)):
for j in range(len(x_grid)):
Z_m[i,j] = len(df[(df['institute']==institutes[i]) &
((df['joining_year']==x_grid[j])) & ((df['gender']=='Male'))])
Z_f[i,j] = len(df[(df['institute']==institutes[i]) &
((df['joining_year']==x_grid[j])) & ((df['gender']=='Female'))])
with plt.xkcd():
fig = plt.figure(figsize=[12,4])
FS = 14
ax = plt.subplot(1,1,1)
clr = 'tab:red'
ax.plot(x_grid,np.sum(Z_m[:,:],axis=0)/(np.sum(Z_f[:,:],axis=0)+1e-3),marker='o',mec='gray',ms=10,alpha=.75,
color='tab:red', label='ALL',linewidth=2)
ax.plot(x_grid,5*np.ones(x_grid.shape),'--',alpha=.75,color='gray',linewidth=2)
# plt.legend(frameon=False, loc='upper right', fontsize=FS)
plt.ylabel('MALE-to-FEMALE FACULTY RATIO\n (only new joinees)',fontsize=FS)
plt.xlabel('YEAR',fontsize=FS)
plt.xticks(fontsize=FS)
plt.yticks(fontsize=FS)
ax.spines['right'].set_visible(False)
ax.spines['top'].set_visible(False)
ax.set_ylim(0,45)
plt.show()
- Before 1993, the ratio was between 10-35, and highly wiggly?
- Something happened after 1995, and the ratio became stable between 5-10.
- Although, the faculty has increased since 2005 (in most institutes), this ratio has not changed much.
Which month do new faculty join?
All of these institutes have a two semester system, namely, July-Dec., and Jan-April. We expect more faculty to join during the start of this semester. This is also seen in the plots at the majority of the institutes. But note, faculty also have been joining round the year.
#collapse
# month of recruitment
x_grid = np.arange(min(df['joining_month']),max(df['joining_month'])+1)
y_grid = np.arange(0,len(institutes))
X, Y = np.meshgrid(x_grid, y_grid)
Z_m = np.zeros(X.shape)
Z_f = np.zeros(X.shape)
for i in range(len(institutes)):
for j in range(len(x_grid)):
Z_m[i,j] = len(df[(df['institute']==institutes[i]) &
((df['joining_month']==x_grid[j])) & ((df['gender']=='Male'))])
Z_f[i,j] = len(df[(df['institute']==institutes[i]) &
((df['joining_month']==x_grid[j])) & ((df['gender']=='Female'))])
with plt.xkcd():
fig = plt.figure(figsize=[16,10])
FS = 10
clr = 'tab:cyan'
for i in range(Z_m.shape[0]):
if i>7:
ax = plt.subplot(3,5,i+3)
else:
ax = plt.subplot(3,5,i+1)
ax.bar(x_grid,Z_m[i,:],color=clr,alpha=1,width=0.35,label='MALE')
ax.bar(x_grid,Z_f[i,:],bottom=Z_m[i,:],color='tab:red',alpha=0.35,width=0.5,label='FEMALE')
plt.legend(frameon=False,loc='upper left', fontsize=FS)
plt.xticks(x_grid,x_grid,rotation=0, fontsize=FS)
ax.grid(color='gray', linestyle='--', linewidth=1,alpha=.3)
plt.ylabel('COUNT',fontsize=FS)
plt.xlabel('MONTH',fontsize=FS)
plt.xticks(fontsize=FS)
plt.yticks(fontsize=FS)
ax.spines['right'].set_visible(False)
ax.spines['top'].set_visible(False)
ax.text(10,max(Z_m[i,:]),institutes[i],fontsize=FS)
plt.show()
#collapse
# day of recruitment
x_grid = np.arange(min(df['joining_day']),max(df['joining_day'])+1)
y_grid = np.arange(0,len(institutes))
X, Y = np.meshgrid(x_grid, y_grid)
Z_m = np.zeros(X.shape)
Z_f = np.zeros(X.shape)
for i in range(len(institutes)):
for j in range(len(x_grid)):
Z_m[i,j] = len(df[(df['institute']==institutes[i]) &
((df['joining_day']==x_grid[j])) & ((df['gender']=='Male'))])
Z_f[i,j] = len(df[(df['institute']==institutes[i]) &
((df['joining_day']==x_grid[j])) & ((df['gender']=='Female'))])
with plt.xkcd():
fig = plt.figure(figsize=[16,20])
for i in range(Z_m.shape[0]):
FS = 10
clr = 'tab:cyan'
if i>7:
ax = plt.subplot(3,5,i+3)
else:
ax = plt.subplot(3,5,i+1)
ax.barh(x_grid,Z_m[i,:],color=clr,alpha=1,height=0.5,label='MALE')
ax.barh(x_grid,Z_f[i,:],color='tab:red',alpha=0.5,height=0.5,label='FEMALE')
plt.legend(frameon=False,loc='upper right')
plt.yticks(x_grid,x_grid,rotation=0, fontsize=FS)
ax.grid(color='gray', linestyle='--', linewidth=1,alpha=.3)
plt.ylabel('COUNT',fontsize=FS)
plt.xlabel('DAY OF MONTH',fontsize=FS)
plt.xticks(fontsize=FS)
plt.yticks(fontsize=FS)
ax.spines['right'].set_visible(False)
ax.spines['top'].set_visible(False)
ax.text(max(Z_m[i,:])//2,10,institutes[i],fontsize=FS+2)
plt.show()
#collapse
vals = []
vals.append(df['months'].values)
vals.append(df['designation'].values)
with plt.xkcd():
fig = plt.subplots(figsize=[12,4])
ax = plt.subplot(1,1,1)
FS = 12
clr = 'cornflowerblue'
sns.distplot(df[df['designation']=='Assistant']['months'],label='Assistant'
,color='tab:red')
sns.distplot(df[df['designation']=='Associate']['months'],label='Associate'
,color='tab:cyan')
sns.distplot(df[df['designation']=='Professor']['months'],label='Professor'
,color='cornflowerblue')
ax.legend(frameon=False, loc='upper right')
ax.vlines(5*12,0,.01,color='k',alpha=0.5)
ax.vlines(10*12,0,.01,color='k',alpha=0.5)
ax.vlines(15*12,0,.01,color='k',alpha=0.5)
ax.text(5*12,0.01,'5 yrs', fontsize=FS)
ax.text(10*12,0.01,'10 yrs', fontsize=FS)
ax.text(15*12,0.01,'15 yrs', fontsize=FS)
ax.grid(color='gray', linestyle='--', linewidth=1,alpha=.3)
plt.xlabel('MONTHS INTO JOB',fontsize=FS)
plt.xticks(fontsize=FS)
plt.yticks(fontsize=FS)
ax.spines['right'].set_visible(False)
ax.spines['top'].set_visible(False)
plt.show()
Assistant Professor age during recruitment ...
For faculty aspirants, a key question is when to apply for a faculty job. Although we like to say - "age is just a number", when it comes to faculty recruitment, usually the age group of new joinees lies in 30-40 years. But there are exceptions too (the youngest Assistant Professor I could notice is in IIT Roorkee, 26 yrs!). Since 2015, we can see there is a peak between 30-40 years.
#collapse
vals = []
years = []
cnt = 5
for i in range(2015,2021):
years.append(i)
vals.append(df[(df['joining_year']==i) & (df['designation']=='Assistant')]['age'].values-cnt)
cnt = cnt-1
with plt.xkcd():
fig = plt.subplots(figsize=[14,6])
FS = 12
clr = 'cornflowerblue'
for i in range(len(years)):
ax = plt.subplot(2,3,i+1)
sns.distplot(vals[i],label=str(years[i]))
ax.legend(frameon=False, loc='upper right')
ax.grid(color='gray', linestyle='--', linewidth=1,alpha=.3)
plt.xlabel('AGE',fontsize=FS)
plt.xticks(fontsize=FS)
plt.yticks(fontsize=FS)
ax.spines['right'].set_visible(False)
ax.spines['top'].set_visible(False)
plt.show()
What about the starting letter of firstnames ...
While I have done basic filtering on names to remove one letter initial, Dr, Prof, SMT, or Mr, etc, still there is a caveat in this observation. All individuals and institutes don't follow the same norm while putting down their name. For example, there are few instances when a surname comes before a first name. But still, lets' see the visualization.
- S is a clear winner!
#collapse
data = {}
keys = ['name', 'first_char']
for key in keys:
data[key] = []
cnt = 0
val_char = []
val_cnt = []
for institute in institutes:
val_char.append([])
val_cnt.append([])
data[keys[0]].append(list(df_fac[df_fac['institute']==institute]['name'].values))
temp = []
for name in data[keys[0]][cnt]:
temp.append(ord(name_filter(name)[0]))
data[keys[1]].append(temp)
for i in range(10):
val_char[cnt].append(chr(max(temp,key=temp.count)))
val_cnt[cnt].append(len(np.where(np.array(temp)==ord(val_char[cnt][i]))[0]))
temp = remove_items(temp, ord(val_char[cnt][i]))
cnt = cnt+1
with plt.xkcd():
fig = plt.subplots(figsize=[12,5])
ax = plt.subplot(1,1,1)
clr = ['wheat','yellowgreen','cornflowerblue']
for i in range(len(institutes)):
for j in range(3):
ax.bar(j+5*i,val_cnt[i][j],color=clr[j], width=1)
ax.text(j+5*i-0.25,val_cnt[i][j],val_char[i][j])
plt.xticks(np.arange(0,61,5)+1,institutes,rotation=45, fontsize=FS)
ax.grid(color='gray', linestyle='--', linewidth=1,alpha=.3)
plt.xticks(fontsize=FS)
plt.yticks(fontsize=FS)
plt.ylabel('COUNT',fontsize=FS)
ax.spines['right'].set_visible(False)
ax.spines['top'].set_visible(False)
plt.show()
That's all for now.
A picture can speak a thousand words. It is really nice to see these institutes documenting and sharing their data. I hope this practice continues, at scale and depth.
- An annual graduating PhD Scholar dataset, detailing thesis title, institute, scholars' age, gender, etc., would be great.
- If you have suggestions or interpretations or corrections ... feel free to comment below or email me
- Together we can continue making our world a better place!
- Please do cross-verify any of your interpretations
Links to the datasheets: IISc, IIT KGP, IIT Bombay, IIT Delhi, IIT Kanpur, IIT-BHU, IIT Roorkee, IIT Mandi, IIT Indore, IIT Jodhpur, IIT Hyderabad, IIT Bhubaneswar, IIT Madras.