Departure Risk

Courses where students who got a DFW were less likely to return the following fall or spring than students who passed — compared to the same pattern among all other students in those courses. This is a correlation, not a cause: the course may reflect a harder structural barrier rather than being the source of attrition. stopout_gap = DFW stop-out rate − pass stop-out rate. Sorted by impact_score = excess gap × DFW count — balances the size of the disproportionate penalty against how many students are affected. Baseline = all non-population students in the same courses. Graduates in their degree term are not counted as stopped out.

Grade Setback

Courses with the highest DFW rates among population students, regardless of whether they predict departure. A high DFW rate is a setback even for students who stay — it delays progress and often triggers a retake. Baseline columns show DFW rates for all non-population students in the same courses.

How These Analyses Work

This page documents exactly how each analysis is computed, derived directly from the source code. Use it to interpret results correctly and spot anomalies.

1. Building a Student Group

A student population is built in three stages: (1) identify candidates — any student who ever appeared in the focal major; (2) classify outcomes — determine what happened to each candidate relative to the major; (3) filter and label — include the desired outcome groups and assign labels. The result (a tibble with student_id, population_label, outcome, entry_pathway, origin, entry_method, entry_status, relevant_until) is passed to every downstream analysis.

Outcomes

• ongoing — still declared in the focal major in the most recent data term.
• graduated — received a degree in the focal major in their last focal term.
• switched_out — left the focal major but remained at UNM. Detected two ways: (1) a formal declaration of another major after their last focal term in cedar_programs; (2) any enrollment record in cedar_students after their last focal term, even without a re-declaration.
• stopped_out — all declared candidates not accounted for by ongoing, graduated, or switched_out. No UNM enrollment or major record after their last focal term.
• chose_elsewhere — appeared only as a pre-major; never declared the focal major, but did declare a different major afterward.
• left_undeclared — appeared only as a pre-major; never declared any major. Left without committing.

Entry pathway (<code>entry_pathway</code>)

How the student arrived at the focal major — computed by get_entry_pathways() :

• direct — first major at UNM was the focal major (no prior declared major or pre-major).
• switched_in — had a non-focal declared major before declaring the focal major.
• pre_major — appeared as a focal pre-major before (or instead of) declaring.

Entry classification columns

• entry_method (classify_entry_method()) — first_program: no prior major record of any kind before this unit; switched_in: had at least one prior major record; unclear: first unit record is at the earliest available term, so prior history is unobservable.
• entry_status (classify_entry_status()) — whether the student’s first record in this unit was as a pre_major or a declared major.

Enrollment window (<code>relevant_until</code>)

Each non-ongoing population student carries a relevant_until term: their last_declared_term (last term with a declared, non-pre-major focal record). Course enrollments after that term are excluded from all analyses. A student who was History for 2 terms, then switched to Business for 8 terms, contributes only the 2 History terms to the analysis. Ongoing students have relevant_until = NA (no restriction).

Worked example — dept = HIST, default scope (declared majors)

student_id	program_name	program_type	is_pre_major	outcome	result
S001	History	Major	FALSE	ongoing	✓ included
S002	History	Second Major	FALSE	ongoing	✓ included (Second Major counts)
S003	History	Major	TRUE	chose_elsewhere / left_undeclared	— excluded by default (pre-major only)
S004	English	Major	FALSE	—	— excluded (different dept)
S005	History	Minor	FALSE	—	— excluded (Minor)

What programs belong to a department? The lookup uses cedar_programs$dept_code, which is populated during transformation via a three-tier lookup: major_dept_map → subj_dept_map → major_code identity. If a program’s dept_code is missing or wrong, it won’t appear in the dropdown.

2. Roadblocks — DFW as a Predictor of Leaving

For each course, compares the fraction of group students who did not return the following term among those who got a DFW grade versus those who passed. The gap between those rates is the key signal.

Step 1: Classify outcomes (per student per course per term)

registration_status_code	final_grade	classified as
DR (early drop)	any	dfw — non-completion regardless of grade
RE / RS / RR	D, D+, D–, F, W, RD, RF	dfw
RE / RS / RR	A–C, CR, P, S, RA–RC, RCR	pass
RE / RS / RR	I, AUD, NR, or other	excluded — ungraded, no signal
WL / other	any	excluded

Step 2: Determine whether each student returned the following term

For each student in each term, we check whether they appear in cedar_students in the next fall or spring. Summer is not counted — skipping summer is normal and not a stop-out.

Graduate correction

Students who earned a degree in term T are not counted as stopped out for that term, even though they don’t appear in term T+1. Without this correction, every graduate who finished their program would be misclassified as a stop-out. The correction uses cedar_degrees$term to identify graduation terms.

Step 3: Compute rates and gap

student_id	BIOL 2310 outcome	returned next term?
S001	pass (A)	yes
S002	pass (B)	no
S003	dfw (F)	yes
S004	dfw (W)	no
S005	dfw (W)	no

pass_stopout_rate = 1/2 = 0.500 (S002 didn’t return)
dfw_stopout_rate = 2/3 = 0.667 (S004, S005 didn’t return)
stopout_gap = 0.667 − 0.500 = 0.167
p_value: chi-squared test on the 2×2 contingency table (outcome × returned). Skipped if either group has fewer than 5 students — result is NA.

4. Course Timing — When Students Take Each Course

Computes the fraction of group students who took each course in their 1st, 2nd, 3rd… enrolled term. Uses relative terms so students who started in different calendar years are aligned on the same axis.

How “term 1” is defined

Relative term 1 is the first term in which the student has a registered course record in cedar_students — not their first semester at UNM, not their first semester in the program, and not any self-reported start date. It is row_number() over their distinct enrolled terms, sorted chronologically by UNM term code.

Skipped semesters

The counter only increments for terms with actual registered enrollment. Gaps are invisible. A student enrolled in Fall, absent in Spring, enrolled in Fall has relative terms 1 and 2 — not 1 and 3. There is no concept of “missed term 2” in this model.

Summer terms

By default, summer does not advance the counter. Summer courses are pinned to the relative term of the immediately preceding fall or spring. A student taking a summer course between their 2nd and 3rd fall/spring semesters has those summer courses recorded as relative term 2. If “Include summer” is enabled, summer gets its own slot in the sequence.

Denominator

For each relative term, the denominator is the number of group students who reached that term — i.e., students whose enrollment record extends to at least that relative term. Students with only 3 terms of data are excluded from relative terms 4–8. This prevents the percentage from being artificially deflated for later terms.

5. Course Pairs — Common Sequences

Finds ordered pairs (A → B) where group students took Course A in one relative term and Course B in a later term, within a configurable term gap.

Exact computation

1. Self-join enrolled records on student_id where term_B > term_A and term_B − term_A ≤ max_term_gap and course_A ≠ course_B.
2. Count distinct students per (course_A, course_B) pair.
3. pct_a_to_b = students who took both ÷ students who took A.

6. Major Changes

Detects when a student’s primary declared major changed from one term to the next, then summarizes those transitions for the selected student group.

Step 1: Detect change events

Source: detect_major_changes() in R/cones/major-changes.R.

1. Filter cedar_programs to program_type == “Major” rows for the population students only.
2. Sort by student_id, term. Use lag() to get each student’s program in the prior term (prev_major) and their prior academic level (prev_level).
3. Flag a change when program_name != prev_major AND (is.na(prev_level) | student_level == prev_level). The level check excludes transitions between undergraduate and graduate programs — a History BA student enrolling in Law School is not a “major change” in the undergraduate sense. is.na(prev_level) passes the first record per student through since there is no prior level to compare.
4. Each flagged row becomes one change event with: student_id, change_term, from_major, to_major, credits_at_change (institutional credits attempted at the time).

Worked example — History student program history

student_id	term	program_name	student_level	prev_major	result
S001	202310	Psychology	Undergraduate	(none)	— first term, no change
S001	202380	Psychology	Undergraduate	Psychology	— same major
S001	202410	History	Undergraduate	Psychology	✓ change event: Psych → History
S001	202480	History	Undergraduate	History	— same major
S001	202710	Juris Doctor	Graduate/GASM	History	— level changed (UG→GR), excluded

Step 2: Derive focal majors

Source: mc_data reactive in R/modules/pathways.R.

Focal majors are the majors that define the selected student group — not all majors ever held by group members. A History cohort student who also declared Political Science should not make PolSci a focal major.

• Dept mode (e.g., HIST): all majors where dept_code == “HIST” and program_type %in% c(“Major”, “Second Major”) in cedar_programs.
• Specific majors mode: exactly the majors the user selected in the population filters.
• Preset mode: the program_names list from the population opt.

Step 3: Filter to focal changes

From the full set of change events, keep only rows where from_major %in% focal_programs OR to_major %in% focal_programs. This means a History cohort sees:

• Psychology → History (arriving to History) ✓
• History → Political Science (leaving History) ✓
• Political Science → Law (made by a History student, but neither side is History) ✗ excluded

Step 4: Build summary outputs

Source: major_change_pathways() in R/cones/major-changes.R.

• Inflow / Outflow table: count distinct to_major (arrivals) and from_major (departures) in focal changes, then filter to rows where the major is in focal_programs. Net = arrivals − departures.
• Common Pathways table: group focal changes by (from_major, to_major), count events, compute avg_credits = average inst_credits_attempted at the moment of the switch. Minimum threshold (default 3) removes rare pairs.
• Avg credits is a proxy for timing: 30 credits ≈ freshman year, 60 ≈ sophomore, 90 ≈ junior. A History → Political Science pair at 75 credits means students are switching in their junior year on average.
• Trend sparkline: per-term count of arrivals (to_major %in% focal, green) and departures (from_major %in% focal, red).
• Donuts: “Leaving for” = top non-focal to_major values among departures. “Arriving from” = top non-focal from_major values among arrivals. Students cycling between focal majors are excluded from the donuts to avoid self-referential loops.

Worked example — Inflow / Outflow for a History dept cohort

major	students arriving to	students leaving for elsewhere	net
History	47	31	+16
History / Pre-Law	5	12	−7

Only History-dept programs appear. The 47 arriving students came from other majors; the 31 departures went to other majors (shown in the “Leaving for” donut).

Methodology reflects: R/branches/population.R (group builder), R/cones/stopout.R, R/cones/pathway.R, R/cones/major-changes.R, and R/modules/pathways.R (display logic). Update this panel when cone logic changes.