M.S. Applied Data Science - Capstone Chronicles 2025

6

Figure 3

listens to what I have to say,” “My supervisor treats me with respect,” and “I know how my work relates to the agency’s goals”. Figure 2 Response Distributions of Survey Questions

Stack Response Rates by Year

4.1.2 Categorical Variables

Several categorical variables were included in the dataset to capture demographic and role-related characteristics of respondents. For instance, the variable dsex represents self-identified sex (coded as ‘A’ for male, ‘B’ for female), while duper indicates supervisory status (‘A’ for supervisors, ‘B’ for non-supervisors). Other variables such as dfendten (tenure with federal agency), drone (remote work status), and dmil (military service background) provide additional contextual insights into the workforce composition. Frequency distributions were computed for each categorical variable. Among all respondents, 56% identified as female (dsex = B) and 44% as male (dsex = A). Approximately 23% of the sample held supervisory roles, while the remaining 77% were nonsupervisory. The drone variable showed 61% of employees reported working remotely either full-time or part-time, whereas 39% were fully on-site. Notably, only 6% of respondents reported having a military service background (dmil = A), which may reflect the specific mission or recruitment focus of the agency. No substantial imbalance was observed in most categorical variables except for supervisory status,

Table 1

Percent Total of Responses by Year

Year 5 2020 5.15% 8.77% 15.14% 40.91% 30.02% 2021 4.40% 6.79% 13.41% 38.77% 36.63% 2022 4.61% 8.83% 15.82% 43.42% 27.32% 2023 4.13% 7.80% 14.80% 43.18% 30.10% 2024 3.94% 7.53% 14.65% 43.01% 30.86% Avg 4.5% 7.94% 14.77% 41.86% 30.99% Figure 3 counts the Likert responses by year. In 2021, responses declined by 53.70% from the prior year. This was a result of the survey being administered later in the year, and the survey being conducted as a sample instead of a census. 1 2 3 4

100