1 Executive Summary

This report presents a preliminary analysis of mathematics course enrollment and outcome data associated with the Emerging Scholar Program (ESP). The combined dataset contains 2,686 enrollment records across multiple mathematics courses and two academic terms in 2024.

ESP participants account for 701 records, or approximately 26.1% of the dataset.

Across courses, ESP participants had a successful completion rate of 85.0%, compared with 74.0% among non-ESP enrollments.

Preliminary regression analysis is used to examine whether this association persists after accounting for course, academic term, demographic characteristics, and available measures of prior academic preparation.

2 Research Questions

This preliminary analysis addresses the following questions:

  • How are ESP participants distributed across mathematics courses and academic terms?
  • Do ESP and non-ESP students differ in rates of successful course completion?
  • Are course outcomes different across individual mathematics courses?
  • Are there demographic differences between ESP and non-ESP enrollment records?
  • Does ESP participation remain associated with successful course completion after accounting for course, term, demographics, and academic preparation?

3 Data and Outcome Definition

The combined dataset represents enrollment records rather than confirmed unique students. Because the available dataset does not contain a de-identified student identifier, multiple enrollments associated with the same individual cannot currently be identified.

For this preliminary analysis, successful completion was defined as follows:

  • For MAT0993, ANC, BNC, and CNC were classified as successful outcomes.
  • UNC and W were classified as unsuccessful outcomes.
  • For other mathematics courses, passing letter grades from A through D- were classified as successful.
  • F and W were classified as unsuccessful.

These classifications should be confirmed with institutional grading and progression policies before final interpretation.

4 Data Completeness

Missing data were handled separately for each analysis rather than excluding records with missing values from the entire dataset.

Missing Data in Key Analysis Variables
total_records Variable Missing Records
2686 Success 47
2686 Esp 0
2686 Course 0
2686 Term 0
2686 Gender 1
2686 Ethnicity 0
2686 Class 0
2686 Act 1421
2686 Sat 946
2686 Test Optional 0

ACT and SAT data contain considerably more missing information than the primary demographic and enrollment variables. For this reason, academic preparation measures are incorporated into separate sensitivity models rather than the primary regression model.

5 ESP Participation

ESP Participation
ESP Status Enrollment Records Percent
Non-ESP 1985 73.9%
ESP 701 26.1%

ESP enrollment records represent approximately 26.1% of the combined dataset.

5.1 Participation by Course

ESP participation varies across courses. This variation is important because course difficulty, student composition, and grading patterns may differ across mathematics courses. Course is therefore included as a control variable in subsequent regression models.

6 Course Outcomes

6.1 Overall Successful Completion

Overall Successful Completion by ESP Status
esp total_enrollments successful_enrollments unsuccessful_enrollments success_rate
Non-ESP 1952 1445 507 74.0%
ESP 687 584 103 85.0%

ESP enrollment records had a successful completion rate of 85.0%, compared with 74.0% among non-ESP records.

This descriptive difference suggests that ESP participation is associated with stronger mathematics course outcomes in the combined dataset. However, descriptive comparisons alone cannot determine whether ESP participation itself accounts for the difference.

6.2 Success by Course

Examining outcomes separately by course helps determine whether the overall ESP difference is broadly observed or concentrated within particular mathematics courses.

6.3 Success by Academic Term

7 Demographic Characteristics

7.1 Gender

Gender Distribution by ESP Status
esp gender n percent
Non-ESP F 900 45.4%
Non-ESP M 955 48.1%
Non-ESP N 129 6.5%
ESP F 267 38.1%
ESP M 372 53.1%
ESP N 62 8.8%

7.2 Ethnicity

Ethnicity Distribution by ESP Status
esp ethnicity n percent
Non-ESP 2 or more races 225 11.3%
Non-ESP American Indian or Alaskan Native 2 0.1%
Non-ESP Asian 442 22.3%
Non-ESP Black or African American 265 13.4%
Non-ESP Hispanic or Latino 146 7.4%
Non-ESP Middle Eastern/North African 284 14.3%
Non-ESP Unknown 13 0.7%
Non-ESP White 608 30.6%
ESP 2 or more races 72 10.3%
ESP American Indian or Alaskan Native 1 0.1%
ESP Asian 124 17.7%
ESP Black or African American 211 30.1%
ESP Hispanic or Latino 95 13.6%
ESP Middle Eastern/North African 63 9.0%
ESP White 135 19.3%

7.3 Academic Class Level

Academic Class Level by ESP Status
esp class_desc n percent
Non-ESP Doctorate 1 0.1%
Non-ESP Freshman 1053 53.0%
Non-ESP Junior 287 14.5%
Non-ESP Post Bachelor 2 0.1%
Non-ESP Senior 139 7.0%
Non-ESP Sophomore 502 25.3%
Non-ESP Unranked Grad 1 0.1%
ESP Freshman 480 68.5%
ESP Junior 56 8.0%
ESP Post Bachelor 1 0.1%
ESP Senior 14 2.0%
ESP Sophomore 150 21.4%

These characteristics are included in the multivariate analysis because differences in the composition of ESP and non-ESP students could contribute to observed differences in course outcomes.

8 Academic Preparation

Available ACT and SAT Indicators by ESP Status
esp total_enrollments act_available act_missing mean_act median_act sat_available sat_missing mean_sat median_sat
Non-ESP 1985 975 1010 22.1 22 1285 700 1111.1 1110
ESP 701 290 411 19.9 20 455 246 1046.0 1050

ACT and SAT scores are included as available proxy measures of prior academic preparation. Because these measures contain substantial missing data, they are evaluated in separate sensitivity analyses.

9 Multivariate Analysis

Logistic regression was used to examine the relationship between ESP participation and successful course completion.

Odds ratios greater than 1 indicate higher odds of successful course completion relative to the reference group.

9.1 Model 1: Unadjusted ESP Association

Model 1: ESP Participation and Course Success
term estimate std.error statistic p.value conf.low conf.high
(Intercept) 2.85 0.0516181 20.290451 <.001 2.58 3.16
espESP 1.99 0.1186823 5.795459 <.001 1.58 2.52

In the unadjusted model, ESP participation was associated with 1.99 times the odds of successful course completion compared with non-ESP participation.

The 95% confidence interval ranged from 1.58 to 2.52.

This association was statistically significant.

This model represents the raw relationship between ESP participation and course success and does not account for differences in course enrollment or student characteristics.

9.2 Model 2: Adjustment for Course and Academic Term

Model 2: Adjusted ESP Association
term estimate std.error statistic p.value conf.low conf.high
espESP 2.74 0.1292396 7.80807 <.001 2.14 3.55

After accounting for mathematics course and academic term, ESP participation was associated with 2.74 times the odds of successful course completion relative to non-ESP participation.

The association remained statistically significant after accounting for course and term.

This model is particularly important because ESP participation is not evenly distributed across courses.

9.3 Model 3: Adjustment for Demographic Characteristics

Model 3: ESP Association After Demographic Adjustment
term estimate std.error statistic p.value conf.low conf.high
espESP 3.19 0.1358998 8.544016 <.001 2.46 4.19

After additionally accounting for gender, ethnicity, and academic class level, ESP participation was associated with 3.19 times the odds of successful course completion.

ESP participation remained a statistically significant predictor of successful course completion in this model.

Because this analysis is observational, the results should be interpreted as evidence of association rather than evidence that ESP participation caused the observed differences.

10 Academic Preparation Sensitivity Analysis

Because ACT and SAT scores contain substantial missing data, models including these variables use smaller subsets of the full dataset.

Number of Enrollment Records Included in Each Regression Model
model n percent_of_dataset
Model 1: ESP only 2639 98.3%
Model 2: ESP + Course + Term 2639 98.3%
Model 3: + Demographics 2638 98.2%
Model 4: + SAT 1709 63.6%
Model 5: + ACT 1236 46.0%
ESP Association Across Logistic Regression Models
model Odds Ratio Lower 95% CI Upper 95% CI P Value
Model 1: ESP only 1.99 1.58 2.52 <.001
Model 2: + Course and Term 2.74 2.14 3.55 <.001
Model 3: + Demographics 3.19 2.46 4.19 <.001
Model 4: + SAT 4.11 2.81 6.12 <.001
Model 5: + ACT 3.18 2.05 5.06 <.001

Changes in the ESP odds ratio across the SAT and ACT models should be interpreted cautiously. These models differ both because an academic preparation variable has been added and because students without the relevant test score are excluded from the analysis.

11 Preliminary Findings

The combined analysis provides several areas for continued investigation:

  • ESP participants demonstrate higher overall successful completion rates than non-ESP students in the mathematics courses represented in this dataset.
  • Differences in ESP participation across courses make course-level adjustment important when estimating the association between ESP and academic outcomes.
  • Logistic regression allows the ESP association to be examined while accounting for course, term, demographic composition, and academic class level.
  • ACT and SAT data provide potentially useful indicators of prior academic preparation but contain substantial missing information.
  • Findings from models using standardized test scores should therefore be treated as sensitivity analyses rather than the primary estimate of ESP’s association with course success.

12 Limitations

This analysis has several important limitations.

First, the dataset contains enrollment records and does not currently contain a de-identified student identifier. As a result, the analysis cannot determine whether individual students appear more than once across courses or terms. Statistical models therefore treat enrollment records as independent observations.

Second, missing data are present for several variables, particularly ACT and SAT scores. Models incorporating these measures use smaller subsets of the dataset and may reflect a different student population from models using the full sample.

Third, this is an observational analysis. While ESP participation may be associated with stronger course outcomes, the analysis does not establish that ESP participation caused those outcomes.

Finally, the current dataset captures course outcomes rather than longer-term educational outcomes such as mathematics progression, STEM persistence, degree completion, graduation, or STEM employment.

13 Recommendations for Continued Analysis

  • Add a de-identified student identifier to support longitudinal and repeated-measures analysis.
  • Confirm the institutional definition of successful completion for each course, particularly whether D-range grades meet progression requirements.
  • Extend the dataset across additional academic years and semesters.
  • Examine persistence into subsequent mathematics and STEM courses.
  • Link ESP participation to major retention, STEM degree completion, and graduation outcomes.
  • Investigate whether ESP outcomes differ across demographic groups.
  • Evaluate whether the association between ESP and course success varies by individual mathematics course.
  • Use alumni interviews to complement quantitative findings and explore ESP’s influence on STEM identity, persistence, and career trajectories.