Topic: Enabling DEI through technology

(1. Sensitization and Awareness)

Team members: Nilakash Paul (Regd No: 211113)

Priyanka Aich (Regd No:221215)
Subrajit Paul (Regd No:221228)

Introduction: In this project our team wanted to create a model that whether we can enabling
DEI through technology and the association between the variables for which we have created
a google form which was circulated among the students who had come with the work
experience who would help us with adequate data for better results.
Traditionally, linear regression modelling in the program Statistical Package for the Social
Sciences (SPSS) is carried out using the REGRESSION procedure, which is capable of fitting
linear models and computing a variety of model fit statistics. However, this aspect of the
program also has limitations:
 Limited to the stepwise method only with no capability of conducting all-possible-subsets
regression
 Limited in terms of optimality statistics for variable selection and existing criteria are in the
form of significance tests prone to Type I/II errors
 Unable to automatically identify and handle outlying cases
 Unable to conduct model ensemble to improve predictions
 Unable to interact with the SPSS Server program to work with very large data.
Given the limitations of the traditional REGRESSION procedure, this article introduces the
new development in SPSS pertaining to linear modelling: The LINEAR procedure which
accelerates the data analysis process through several automatic mechanisms.

Objectives: To understand how technology is playing role in removing biasness of diversity,

equity and inclusion.
Our team tried to take out the central tendency of the scale variables found out in the survey,
whose inferences are found out respectively:
Inferences: The data set contains 104 respondents' comments regarding their job
environment and how much they believe their identity is accepted and valued. All of the
variables fall within a range of 4, with a minimum score of 1 and a highest score of 5. With a
standard deviation of 1.259, the mean result for "bringing your whole self to work regardless
of the background or identity" is 3.08. It differs by 1.586. The distribution of scores is shown
to be comparatively flat when compared to a normal distribution by the kurtosis value of -
1.193. With a standard deviation of 1.191, the mean score for "how well does the
organisation accommodate employees with different backgrounds or identities" is 3.19. The
variance is 1.419. The kurtosis value of -1.007 also indicates that the distribution of scores is
relatively flat compared to a normal distribution. The mean score for "how well does your
organization address issues of discrimination or bias in the workplace" is 3.02, with a
standard deviation of 1.123. The variance is 1.262. The negative kurtosis value of -0.864
suggests that the distribution of scores may be slightly skewed towards the lower end. The
mean score for "how well does the organization communicate its commitment to DEI to
employees" is 3.06, with a standard deviation of 1.113. The variance is 1.239. The kurtosis
value of -0.690 indicates that the distribution of scores is relatively normal. The mean score
for "satisfaction with opportunities for growth and advancement within your organization
regardless of your background or identity" is 3.04, with a standard deviation of 1.146. The
variance is 1.312. The negative kurtosis value of -0.951 suggests that the distribution of
scores may be slightly skewed towards the lower end. The valid N for all the variables is 103,
indicating that there are missing data points for one of the variables. Overall, the data
suggests that there is room for improvement in the workplace with regards to diversity,
equity, and inclusion.

In this case processing summary, the valid number of respondents is 104 and there is no
missing respondent in this processing summary where the best describes your gender identity
and racial or ethnic groups.
The following interpretation is for the gender identity and the racial or ethnic groups the
following outputs are as follows the Asian is 45, Black African American 0, Indian is 1,
white is 2, white and Asian is 0 and total is 50 in female and in male is total is 54, total is
104.

Inferences: The Pearson Chi-Square value is 7.535, with 7 degrees of freedom and an
asymptotic significance (p-value) of 0.375. The likelihood ratio value is 10.241, with 7
degrees of freedom. The fact that 14 cells (87.5%) have expected counts less than 5, with the
minimum expected count being 0.48, indicates that the sample size may be too small or that
the distribution of the data is too skewed. This can result in inaccurate test results and
conclusions. When conducting a Chi-Square test, it is important to ensure that the
assumptions of the test are met, including having adequate sample sizes in each cell and
having a roughly equal distribution of observed frequencies across the cells. If these
assumptions are not met, it may be necessary to use a different statistical test or to adjust the
data to meet the assumptions.
The "Variables Entered" section indicates that only one independent variable was included in
the model: "On a scale, how satisfied are you with the opportunities for growth and
advancement within your organization regardless of your background or identity?" The
"Variables Removed" section is blank, indicating that no independent variables were
removed from the model. The "Method" used in this analysis was "Enter", which means that
all the variables were entered simultaneously into the model. The "Dependent Variable" is
"On a scale of, how comfortable do you feel bringing your whole self to work regardless of
your background or identity?” Overall, this output suggests that the analysis focused on
examining the relationship between how satisfied individuals were with growth and
advancement opportunities in their organization, and how comfortable they felt bringing their
whole selves to work regardless of their background or identity.

The "Model Summary" section provides information about the overall fit of the model. The R
square value is 0.690, which indicates that approximately 69% of the variation in the
dependent variable can be explained by the independent variable. The adjusted R square
value is 0.476, which suggests that the model may not be the best fit for the data. The
"Change Statistics" section provides information about the change in the R square value
when the independent variable is added to the model. The R square change is 0.690,
indicating that the independent variable explains a significant amount of the variation in the
dependent variable. The F change value is 91.592 with 1 and 470 degrees of freedom, and the
p-value (Sig. F Change) is 0.000, which indicates that the model is statistically significant.
Overall, this output suggests that there is a strong relationship between how satisfied
individuals are with growth and advancement opportunities in their organization, and how
comfortable they feel bringing their whole selves to work regardless of their background or
identity. However, the adjusted R square value suggests that there may be other variables that
should be considered in the model to better explain the variation in the dependent variable.

The "Coefficients" section provides information about the regression coefficients. The
unstandardized regression coefficient (B) for the independent variable "On a scale, how
satisfied are you with the opportunities for growth and advancement within your organization
regardless of your background or identity?" is 0.255. This suggests that for every one unit
increase in the independent variable, the dependent variable ("On a scale of, how comfortable
do you feel bringing your whole self to work regardless of your background or identity?")
increases by 0.255 units. The "Standardized Coefficients" section provides information about
the beta coefficient, which indicates the standardized effect of the independent variable on the
dependent variable. The beta coefficient for the independent variable is not provided in this
output. The "t" value is 3.188, which indicates that the coefficient is statistically significant.
The "Sig." value is 0.002, which suggests that the coefficient is statistically significant at the
p < 0.05 level. Overall, this output suggests that there is a statistically significant positive
relationship between how satisfied individuals are with growth and advancement
opportunities in their organization, and how comfortable they feel bringing their whole selves
to work regardless of their background or identity.
The "Model" row provides information about the regression model. The "Regression Sum of
Squares" is 75.630, which represents the amount of variation in the dependent variable that is
explained by the independent variable. The "df" (degrees of freedom) for the model is 1,
indicating there is one independent variable. The "Mean Square" is also 75.630, which is the
sum of squares divided by the degrees of freedom. The "F" value is 91.592, which is
calculated by dividing the regression mean square by the residual mean square. This provides
information about the ratio of explained variance to unexplained variance in the model. The
"Sig." value is 0.000, which suggests that the model is statistically significant at the p < 0.05
level. The "Residual" row provides information about the unexplained variance in the model.
The "Sum of Squares" is 83.399, and the "df" is 102, indicating the residual degrees of
freedom. The "Total" row provides information about the total variance in the dependent
variable. The "Sum of Squares" is 159.029, and the "df" is 101, indicating the total degrees of
freedom. Overall, this output suggests that the regression model is statistically significant and
that the independent variable ("On a scale, how satisfied are you with the opportunities for
growth and advancement within your organization regardless of your background or
identity?") significantly predicts the dependent variable ("On a scale of, how comfortable do
you feel bringing your whole self to work regardless of your background or identity?"). The
regression model explains a significant amount of the variation in the dependent variable.

Summary and findings:

1. The descriptive test was performed to find out whether their needs to be an
improvement in the workplace so the test gave that the model’s data suggested that
there is room for improvement in the workplace with regards to diversity, equity, and
inclusion.
2. The chi square test was performed to find out the association whether there is a
association between the racial groups to be identified and which describes the best
gender identity While performing the chi square test its results suggested that there is
no significant association between the variables, but the violation of the chi square
test assumptions due to small expected cell count raises concerns about the validity of
the test.
3. While performing regression analysis, the model’s output suggested that there is a
strong relationship between how satisfied individuals are with growth and
 advancement opportunities in their organization, and how comfortable they feel
bringing their whole selves to work regardless of their background or identity.
4. The model’s output suggests that there is a statistically significant positive
relationship between how satisfied individuals are with growth and advancement
opportunities in their organization, and how comfortable they feel bringing their
whole selves to work regardless of their background or identity
Automatic Variable selection
Among many variable selection methods, the stepwise method and the all-possible-subsets
(i.e., best-subsets) methods remain as popular techniques thanks to their availability in major
statistical computer programs; although, certain regularization methods like the Least
Absolute Shrinkage Selection Operator or LASSO by Tibshirani (1996) are gradually taking
over. The two popular approaches differ primarily in the extent to which the model space is
searched. The stepwise approach searches only a subset of the entire model space, whereas
the all-possible-subsets approach explores the space in its entirety. Also, because selecting
variable subsets requires a numerical criterion that measures the quality of each subset, both
approaches count on many of the same optimality criteria.
After a criterion is specified, each approach iteratively searches for the best predictor
subset(s) that optimize the criterion.
Next, the two automatic search methods are discussed in greater detail:
 Stepwise method: This approach enters or removes predictors one at a time after taking into
account the marginal contribution of a predictor controlling for other variables already in the
model. Multiple criteria exist for evaluating this marginal contribution: the Partial F-test, the
Akaike’s Information Criterion Corrected (AICC) (Hurvich & Tsai, 1989), etc.
The stepwise method has three variants: 1) forward selection, 2) backward elimination, and
3) forward stepwise (Efroymson, 1960).
The forward stepwise variant is probably the most popular technique, where after each
variable (other than the first) is added to the set of selected predictors, the algorithm examines
to see if any of the previously selected variables could be dropped without appreciably
increasing the residual sum of squares. This process continues until the stopping criteria are
met.
 All-possible-subsets method: Compared with the stepwise approach that economizes on
computational efforts by exploring only a certain part of the model space, the all-possible
subsets approach conducts a computationally intensive search of the entire model space by
considering all possible regression models from the pool of potential predictors. Given p
predictors, there is a total of 2 p regression models (including the intercept-only model) to be
estimated. Clearly, the number of models under the all-possible-subsets approach increases
rapidly as the number of candidate predictors increases. Given that the approach is
computationally intensive, it works better when the number of potential predictors is not too
large, for example 20 or fewer.
Data preparation
The data was prepared and collected by the method of primary data collection by preparing a
questionnaire and circulating among the students who have work experience who would be
better to relate with the pros and cons of the workplace.
Sample size: 104
Variables taken: All scale variables are taken for the convenience of building the model.
Fields: On a scale of 1-5, how comfortable do you feel bringing your whole self to work,
regardless of your background or identity?
On a scale of 1-5, how effectively does your organization promote diversity and inclusivity in
the workplace?
On a scale of 1-5, how satisfied are you with the opportunities for growth and advancement
within your organization regardless of your background or identity?
On a scale of 1-5, how well does your organization communicate its commitment to DEI to
employees?
On a scale of 1-5, how well does your organization accommodate employees with different
backgrounds or identities?
Target variable: On a scale of 1-5, how well does your organization address issues of
discrimination or bias in the workplace?
Predictors:
Name:
What is your age range?
Which of the following best describes your gender identity?
Which of the following racial or ethnic groups do you identify with?
Do you feel your organization values diversity?
Has your organization implemented any DEI initiatives or programs?
Have you experienced any discrimination or bias at your workplace?
Has your organization provided any DEI training or education to employees?
Column row V3
Merge categories to maximize association with
Actions Taken (Race_transformed)
target
Role (Age_transformed) Predictor
Merge categories to maximize association with
Actions Taken (Age_transformed)
target
Role (Training_transformed) Predictor
Role (Diversity_transformed) Predictor
Merge categories to maximize association with
Actions Taken (Training_transformed)
target
Merge categories to maximize association with
Actions Taken (Diversity_transformed)
target
Role (Race_transformed) Predictor
Discussion:
A positive coefficient indicates that as the value of the independent variable increases,
the mean of the dependent variable also tends to increase. A negative coefficient suggests
that as the independent variable increases, the dependent variable tends to decrease.
The model building summary is a forward stepwise using the information criterion.
Checkmark shows that the effect in the model at this step.
The model accuracy came to be 25% which shows that the target respondents were less so the
model fits better in this range.
While calculating the estimated means, the target variable taken was on a scale how well does
your organization address issues of discrimination or bias in the workplace?
The significant effects (p<0.05) is displayed:
We could see a steep slope while comparing it with the DEI training or education to
employees.
We could see a steep slope while comparing it with your organization value diversity.
We could see a variation in slope while comparing it with the age range.
While checking the automatic data preparation: the field taken into account were the age
transformation, diversity transformation, race transformation, training transformation. And
role taken into account is the Predictor and actions which were taken were merged categories
to maximize association with the target variable.
While checking the predictor importance with the target variable:
The result found that there was a maximum level of age range preceding the training given to
the employees in DEI than the organization value diversity or not.
The predicted by observed value found was how well does your organization address issues
of discrimination or bias in a workplace the counts were found to be 20,15,10,5,0; with
maximum counts in 10.
While comparing the residuals the histogram of studentized residuals compares the
distribution of the residuals to a normal distribution. The smooth line represents the normal
distribution. The closer the frequencies of the residuals are to this line, the closer the
distribution of the residuals is to the normal distribution.
The outliers found while comparing the target variable, we used the Cooks distance to be
0.109 which shows low influence in the model computations, which distort the model
accuracy.

Limitations:
Sample size is small and the model is not very strong in enough to support any of these
claims of the above study.

CONCLUSION:
Regarding the research title "DEI using technology," it is evident that the use of technology
has great potential for promoting diversity, equity, and inclusion in various areas. Still, it is
essential to note that the implementation of technology in DEI initiatives should be done in a
thoughtful and intentional way to ensure that it does not perpetuate discrimination or
reinforce existing inequalities.
Organizations need to assess whether they value diversity and how much they prioritize it in
their operations. It is also crucial for organizations to implement DEI initiatives or programs
and provide training or education to their employees to promote a culture of inclusion. This
will not only help employees understand and embrace diversity but also provide them with
the skills and tools they need to identify and address discriminatory behaviours and attitudes.
Finally, it is important to acknowledge that discrimination and bias still occur in the
workplace, and organizations must take steps to address and eliminate such behaviours. By
implementing DEI initiatives, providing training and education, and leveraging technology in
a thoughtful and inclusive way, organizations can create a workplace culture that values
diversity and promotes equity and inclusion for all employees.