You are on page 1of 3

MEMO

to : Director Of Human Resources

from : Nguyen Thi Mai Chi

subject : Accessing The Variables Of Employee’s Salary By Processing The Regression


Model And Dummy Coding

date : November 19th, 2022

In light of the recent mandate requiring the compilation of annual salary reports for the
upcoming fiscal period and the concurrent exploration of employing a prediction model
to anticipate wage levels for each employee, I am drafting this memorandum to undertake
a thorough investigation into the intricate relationship between predictor variables and the
salary amounts allocated to individual employees.
Initially, the dataset comprised 8378 samples and six variables. However, due to the
inherent limitations of categorical data for regression modeling, the adoption of a
"dummy coding" approach became imperative. This method involved extracting
categorical variables from each category variable, resulting in a refined dataset
encompassing nine predictor factors: AGE, White, Black, American Indian, Asian or
Pacific Islander, Gender (Male), Belong to Union, Education, and Wage.
For a more nuanced understanding, let's delve into the intricate analysis and interpretation
of the relationship between these nine predictor variables (x-values) and the salaries of
faculty members (y-value).
The derived prediction formula for salary is as follows:
Wage=
-7.52039+0.107646*Age+0.912176*Education+1.92665*Gender+0.651962*White-
0.26068*Black+0.739851*American Indian+0.38354*Asian or Pacific
Islander+2.954559*Belong to Union
In essence, the model operates such that when additional x-values are introduced, the
projected y-value experiences an increase or decrease corresponding to the coefficient of
each determinant, while all other variables remain constant. For instance, an employee's
pay increases by 0.1 for each additional year of Age, assuming all other factors remain
constant. Conversely, holding all other variables constant, an employee's pay decreases
by 0.2606 for every additional Black individual.
Examining the coefficients reveals discernible patterns of discrimination among
employee races. Wages for American Indians and Whites consistently trend higher,
whereas wages for Asians and Pacific Islanders consistently trend lower. Notably, the
negative coefficient associated with the Black variable indicates that Black employees
receive the lowest wages, indicative of a concerning disparity. Additionally, the model
underscores significant gender-based wage inequality, with men earning substantially
more than women. This observation serves as a potent demonstration of the
discriminatory and inequitable nature of the existing compensation policy.
On a more positive note, both belonging to a union and possessing a higher level of
education exhibit positive coefficients. These findings align with expectations, as union
members typically earn more than their non-union counterparts, and higher educational
attainment corresponds to higher wages.
However, it's imperative to note that the Adjusted R^2of the model is only 0.235,
significantly below 1. This suggests that the model's efficacy is limited, prompting the
need to explore alternative methods for accurately estimating each employee's salary
level.
Best regards,
Nguyen Thi Mai Chi

You might also like