Analysis for POD's Internship Report.

POD UW collaborated with the Association for Computing Machinery (ACM) to survey students from the Paul Allen School of Computer Science and released a report about trends and correlations; See the full report here

This is the second part of my work after Cleaning and processing the data

Here are the visualizations that ended up on the report:

The Code

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import re

rawResults = pd.read_csv("ACMExtended.csv")

rawResults.head()

	Unnamed: 0	Timestamp	expectedGraduation	priorInternship	returnOffer	rolesAppliedFor	firstApplication	lastApplicatoin	numAppliedTo	resumeSup	...	bigTechAssessment	bigTechOffers	avgIntvPrepTime	avgIntvPrepSpan	avgIntvPrepHrs	numOffers	avgInitialReply	avgAssessmentToFirstIntvTime	avgFirstToFinalIntvTime	avgIntvToOfferTime
0	0	11/25/2020 15:58	6/11/2022	Yes	No	Software Engineer Internship	7/1/2020	8/15/2020	150	LinkedIn, Personal GitHub, School / class proj...	...	1	1	10.0	10.0	40.0	3	3.0	1.5	2.500000	1.0
1	1	11/25/2020 16:15	6/10/2023	No	NaN	Software Engineer Internship, 1st / 2nd year i...	8/28/2020	10/30/2020	36	Personal projects	...	1	0	1.5	1.5	4.0	1	0.0	NaN	1.714286	1.0
2	2	11/25/2020 16:25	12/17/2022	No	NaN	Software Engineer Internship	8/26/2020	10/16/2020	165	LinkedIn, Personal GitHub, Personal website, P...	...	1	1	3.5	3.5	13.0	3	2.0	2.0	1.000000	1.0
3	3	11/25/2020 16:37	6/14/2022	Yes	Yes	Software Engineer Internship	8/20/2020	11/4/2020	70	LinkedIn, Personal GitHub, Personal projects	...	1	0	3.5	3.5	1.0	1	1.0	NaN	NaN	NaN
4	4	11/25/2020 17:18	3/31/2022	No	NaN	Software Engineer Internship	7/10/2020	11/19/2020	270	LinkedIn, Personal GitHub, Personal projects, ...	...	0	0	10.0	10.0	8.0	4	NaN	NaN	NaN	NaN

5 rows × 60 columns

# A quick look to see how many offers respondents received
rawResults.numOffers.value_counts()

0    37
1    33
2    25
3     7
4     6
5     2
6     1
Name: numOffers, dtype: int64

# Stratifying respondents that recieved no offers and 1+ offers
rawResults["gotOffer"] = np.zeros(rawResults.shape[0])
rawResults.loc[rawResults.numOffers > 0, "gotOffer"] = 1
offersN = rawResults.gotOffer.value_counts()
offersN

1.0    74
0.0    37
Name: gotOffer, dtype: int64

totalN = rawResults.shape[0]
totalN

111

offersVsNon = rawResults.groupby("gotOffer")

##TODO MOVE

rawResults.groupby("bigTechOffers")[["avgIntvPrepHrs", "avgIntvPrepSpan"]].mean()

	avgIntvPrepHrs	avgIntvPrepSpan
bigTechOffers
0	4.838028	9.852113
1	7.525000	9.250000

Average prep time for respondents that received an offer vs respondents with no offers

round(offersVsNon[["avgIntvPrepHrs", "avgIntvPrepSpan"]].mean(), 2)

	avgIntvPrepHrs	avgIntvPrepSpan
gotOffer
0.0	4.99	7.41
1.0	6.22	10.75

offersVsNon[["avgIntvPrepHrs", "avgIntvPrepSpan"]].median()

	avgIntvPrepHrs	avgIntvPrepSpan
gotOffer
0.0	1.0	3.5
1.0	4.0	10.0

Comparing which portfolio documents were included the most

supplimentary = offersVsNon[["inclGithub", "inclPersonalProj", "inclSchoolProj", "inclWebsite", "inclLinkedIn"]]
supplimentarySums = supplimentary.sum()
supplimentaryPct = supplimentary.mean()

round(supplimentaryPct, 2) #Percent from offered (vs non offered) applicants included the following 
# Percent from specific sample pool

	inclGithub	inclPersonalProj	inclSchoolProj	inclWebsite	inclLinkedIn
gotOffer
0.0	0.46	0.57	0.54	0.03	0.89
1.0	0.55	0.62	0.47	0.28	0.88

round(supplimentarySums/totalN*100, 2) # Percent from overall sample pool

	inclGithub	inclPersonalProj	inclSchoolProj	inclWebsite	inclLinkedIn
gotOffer
0.0	15.32	18.92	18.02	0.90	29.73
1.0	36.94	41.44	31.53	18.92	58.56

rawResults.groupby("inclWebsite").gotOffer.mean() # percent of applicants who received offers (website included vs not)

inclWebsite
0    0.595506
1    0.954545
Name: gotOffer, dtype: float64

95% of applicants that included a website on their applications also received at least 1 offer.

supplimentary = rawResults.groupby("bigTechOffers")[["inclGithub", "inclPersonalProj", "inclSchoolProj", "inclWebsite", "inclLinkedIn"]]
btSums = supplimentary.sum()
btPct = supplimentary.mean()

btPct

	inclGithub	inclPersonalProj	inclSchoolProj	inclWebsite	inclLinkedIn
bigTechOffers
0	0.535211	0.619718	0.535211	0.15493	0.859155
1	0.500000	0.575000	0.425000	0.27500	0.925000

Comparing interview prep resources

# Sum of each perp resource used.
prep = rawResults[["gotOffer", "leetCodePrep", "crackingTheCodePrep", "youtubePrep", "mockInterviewPrep", "hackerrankPrep", "crackingThePmPrep", "492Prep"]]
prep.sum()

gotOffer                74.0
leetCodePrep           105.0
crackingTheCodePrep     57.0
youtubePrep             44.0
mockInterviewPrep       39.0
hackerrankPrep          39.0
crackingThePmPrep        1.0
492Prep                  1.0
dtype: float64

round(prep.groupby("gotOffer").mean()*100, 2) # percent of respective sample used each resource.

	leetCodePrep	crackingTheCodePrep	youtubePrep	mockInterviewPrep	hackerrankPrep	crackingThePmPrep	492Prep
gotOffer
0.0	91.89	56.76	40.54	40.54	40.54	0.00	0.00
1.0	95.95	48.65	39.19	32.43	32.43	1.35	1.35

round(prep.leetCodePrep.mean(), 2) # LeetCode usage percent from total sample

0.95

def hourGrp(hrs): # Defining 4 main hour groups for visualization
    if (hrs < 4):
        return "1-3"
    elif (hrs < 8):
        return "4-7"
    elif (hrs < 15):
        return "8-14"
    else:
        return "15+"
rawResults["prepHoursGrp"] = rawResults.apply(lambda x: hourGrp(x.avgIntvPrepHrs), axis=1)

rawResults[["avgIntvPrepHrs", "prepHoursGrp"]]

	avgIntvPrepHrs	prepHoursGrp
0	40.0	15+
1	4.0	4-7
2	13.0	8-14
3	1.0	1-3
4	8.0	8-14
...	...	...
106	8.0	8-14
107	4.0	4-7
108	4.0	4-7
109	1.0	1-3
110	4.0	4-7

111 rows × 2 columns

import statsmodels.formula.api as smf

re.findall("\d+", "more than 50")[0]

'50'

def refferals(x):
    try:
        num = re.findall("\d+", x)
        return int(num[0])
    except:
        print("failed to convert ", x)
rawResults["numRefferalsInt"] = rawResults.apply(lambda x: refferals(x.numReferrals), axis=1)

failed to convert  nan
failed to convert  nan
failed to convert  Three
failed to convert  nan
failed to convert  nan
failed to convert  nan
failed to convert  nan
failed to convert  nan
failed to convert  nan
failed to convert  nan
failed to convert  nan
failed to convert  nan
failed to convert  nan
failed to convert  nan
failed to convert  nan
failed to convert  nan
failed to convert  nan
failed to convert  nan
failed to convert  nan
failed to convert  nan
failed to convert  nan
failed to convert  nan
failed to convert  nan
failed to convert  nan
failed to convert  nan
failed to convert  nan
failed to convert  nan
failed to convert  nan
failed to convert  nan
failed to convert  nan

rawResults["nAppliedTo"] = rawResults.apply(lambda x: refferals(x.numAppliedTo), axis=1)

test = smf.ols(formula="numOffers ~ avgIntvPrepSpan", data=rawResults)
mod = test.fit()
mod.summary()

OLS Regression Results

Dep. Variable:	numOffers	R-squared:	0.005
Model:	OLS	Adj. R-squared:	-0.004
Method:	Least Squares	F-statistic:	0.5414
Date:	Sun, 24 Jan 2021	Prob (F-statistic):	0.463
Time:	11:19:36	Log-Likelihood:	-187.40
No. Observations:	111	AIC:	378.8
Df Residuals:	109	BIC:	384.2
Df Model:	1
Covariance Type:	nonrobust

	coef	std err	t	P>|t|	[0.025	0.975]
Intercept	1.2175	0.166	7.342	0.000	0.889	1.546
avgIntvPrepSpan	0.0083	0.011	0.736	0.463	-0.014	0.031

Omnibus:	24.151	Durbin-Watson:	2.030
Prob(Omnibus):	0.000	Jarque-Bera (JB):	31.654
Skew:	1.166	Prob(JB):	1.34e-07
Kurtosis:	4.188	Cond. No.	19.5

Notes:
[1] Standard Errors assume that the covariance matrix of the errors is correctly specified.

rawResults[["gotOffer", "prepHoursGrp", "avgIntvPrepHrs", "avgIntvPrepSpan"]].to_csv("offersAndPrepTime.csv")