Juan Camilo Gomez R - 201423604

Case Analysis

1. How much money can students expect to make on graduation?

For this question, we compute the mean or the expected value of salaries for all respondents.
Data was filtered so the respondents who didn’t answer or did not disclose (codes 998 and 999)
were not considered.

The mean is as follows:

So, students can expect on graduation a salary of around 54985.32 upon graduation.

2. Are there any variables (for example age, gender, quartile, language spoken or work
experience, etc) that affect how much a student can expect to make?

Variables that can be expected to affect how much a student can make are age, sex, quartile
ranking (which is best to summarize grades), years of work experience, and first language.

GMAT scores and average marks are not considered as it is considered that quartile ranking can
best describe the academic performance of each student. It is also inferred that satisfaction
does not have any relation with salary.

We then conduct a linear regression analysis to determine the dependence of the dependent
variable, salary, on one or more other variables, the explanatory variables (independent
variables), age, sex quartile, experience, and language. We conduct regression analyses as a
model and then one by one to determine if they affect or not how much a student can expect
to make. We use alpha p=0.05 to determine the significance.
Juan Camilo Gomez R - 201423604

Model

Linear regression was performed to quantify the relationship between independent variables
(seen in the first column) and salary (dependent variable). A sample of 193 respondents was
used in the analysis.

Here we can see that the only variables that are significant considering the whole model are
age and quarter with a p-value of 0.019 and 0.04. Nevertheless, it is a weak model with an Adj-
R squared of only 0.0304, which is the value that the variables explain 3.04% of the change in
salary.

Age and Salary

Linear regression was performed to quantify the relationship between age and salary. A sample
of 193 respondents was used in the analysis.
 
Results showed that there was a statistically insignificant relationship between age and salary
(p = 0.071 > 0.05) and age accounted for 1.18% of explained variability in salary. 
Juan Camilo Gomez R - 201423604

  
Each additional year was associated with a decrease, on average, of -1640.948 monetary units.

Ultimately considering the model the variable becomes significant (see the first model)

The contribution or impact that the variable has on the model is the difference between the adj
r squared with the variable considered and not considered which is 0.0304 – 0.0274 (new
model) which is 0.003 of impact on salary.

Gender and Salary

Linear regression was performed to quantify the relationship between sex and salary. A sample
of 193 respondents was used in the analysis.
 
Results showed that there was a statistically insignificant relationship between age and salary
(p = 0.798 > 0.05) and sex accounted for 0.49% of explained variability in salary. 

Quartile and Salary

Linear regression was performed to quantify the relationship between quarter and salary. A
sample of 193 respondents was used in the analysis.
 
Results showed that there was a statistically significant relationship between quarter and salary
(p = 0.041 < 0.05) and quarter accounted for 1.66% of explained variability in salary. 
Juan Camilo Gomez R - 201423604

  
As quartiles moved up it was associated with a decrease, on average, of -7088.862 monetary
units.

Considering the first model quartile is still significant both with all variables considered and
alone. The contribution that quartile has on the whole model is the Adj r squared of the whole
model minus the adj r squared without quartile, which is 0.0304 – 0.0135 which is 0.0169 that
accounts for 1.69% of the change in salary (weak relationship).

Work experience and salary

Linear regression was performed to quantify the relationship between work_yrs and salary. A
sample of 193 respondents was used in the analysis.
 
Results showed that there was a statistically insignificant relationship between work_yrs and
salary (p = 0.462 > 0.05) and work_yrs accounted for 0.24% of explained variability in salary. 
  
As work years increased it was associated with a decrease, on average, of -767.6263 monetary
units.

First language and Salary

Linear regression was performed to quantify the relationship between frstlang and salary. A
sample of 193 respondents was used in the analysis.
 
Results showed that there was a statistically insignificant relationship between frstlang and
salary (p = 0.922 > 0.05) and frstlang accounted for -0.52% of explained variability in salary. 

We then conclude that seeing the model, age and quarter are the only ones that do affect in a
low form (less than 5 % change) the salary that students can expect to make, it is a very weak
model with Adj-r squared of 3.04% of the change in salary attributed to the independent
variables. On their own only the variable of the quartile can affect how much a student can
expect to make.

3. Are the salary data figures published by MBA programs believable?

Seeing the results of the linear regression made, the first table (Model), and the only variables
that affect (quartile and age) in such a weak way makes us think about the credibility of data. As
logic and theory suggest with age the salary level increases in a strong relationship, this as well
as years of work experience, and first language; we would expect that the Adj r squared in the
model would be greater and would account for way much more % of the change in the level of
salary, being a strong model (more than 0.7). Having those weak relationships, we might think
that the salary figures given were randomized or that do not account for the reality of what a
student can or will make. Since there is no variable that would explain with a strong
relationship and thoroughly why some students make more than others, we conclude that the
salary data figures published by the MBA program are not believable.