Multivariate Analysis: Drivers of Confidence in Tech Skills

What drives girls' confidence in tech skills?

Multivariate analysis based solely on the provided survey dataset.

Pre-stage low confidence
100% (5/5)
Mid-stage average confidence
2.33/4
Score ↔ Confidence (Mid)
~0.00 r
Built a web app (Mid)
66.7% (4/6)
Improved Pre → Mid
4 of 5
Notes: Confidence scoring mapped as Not=1, Somewhat=2, Confident=3, Very=4 (used for averages and correlation).
Confidence legend: Very confident Confident Somewhat confident Not confident

Confidence by stage

Pre vs Mid (based on 5 Pre and 6 Mid responses with confidence statements).

Pre-stage
n=5
100% Not confident
Mid-stage
n=6
Shift from universally low confidence (Pre) to a mixed profile (Mid): 33% at confident/very, 50% somewhat, 17% not.

Individual confidence transitions (Pre → Mid)

Based on participants with entries at both stages.

  • Rosura: Not confident → Very confident (built multiple web apps; explicit mentor feedback)
  • Enas: Not confident → Somewhat confident (built a web app; still struggles with complex problems)
  • Irina: Not confident → Somewhat confident (built a web app; basic coding OK, complex still hard)
  • Triana: Anxious → Confident (no web app; solves challenges; peer motivation)
  • Ying: Not confident → Not very confident (despite 92 score and a web app; debugging is a barrier)
Net: 4 of 5 improved; 1 remained low due to debugging difficulties.

Factor 1: Project experience (building a web app)

Mid-stage only: 4 built a web app; 2 did not.

Built a web app (n=4)
Outcomes vary; projects help but are not sufficient alone.
Did not build a web app (n=2)
Confidence can rise via practice/problem-solving even without a full app (Triana).
Signal: The strongest jump to "very confident" coincides with building multiple apps plus explicit mentor feedback (Rosura).

Factor 2: Test scores

Among Mid participants, correlation between test score and confidence is approximately 0.00 (no linear relationship observed).

Low scores (≤44, n=2)
Avg confidence: 2.0
Mid scores (65–75, n=2)
Avg confidence: 3.0
High scores (≥85, n=2)
Avg confidence: 2.0

Implication: High scores do not guarantee confidence (e.g., Ying scored 92 but reports low confidence due to debugging issues).

Factor 3: Mentorship and feedback

The only participant explicitly citing mentor feedback (Rosura) is the sole "very confident" respondent at Mid. This suggests targeted, affirming feedback tied to shipped projects may amplify confidence gains.

“I feel very confident... built a few web apps from scratch, and my mentors have given me positive feedback.”

Factor 4: Debugging proficiency

Debugging barriers suppress confidence even with strong test performance and shipped projects (Ying: score 92, web app built, still not confident due to debugging).

Confidence drag indicator: “I often need help debugging my code and don't always understand what went wrong.”

Factor 5: Consistency & attendance

Inconsistent participation correlates with feeling overwhelmed and lower advocacy (Pravalika: Somewhat confident; NPS 4).

“I missed a lot of sessions... often feel overwhelmed.”

Factor 6: Peer community & challenge-based practice

Confidence can rise without a complete app when regular problem-solving and peer support are present (Triana: Confident; no app; emphasizes challenges and friendships).

“I feel motivated and confident... built friendships with girls that have the same interests.”

Resume workshop and long-term outcomes

All Mid-stage respondents attended the resume workshop; many reported interview boosts. While this is not a direct coding-confidence driver, the positive external validation may reinforce overall self-efficacy. Post-stage, Rosura reports sustained impact and career progression.

“Landed several interviews because of my new resume.” — Rosura (Mid)
“Game-changer for my job search.” — Ying (Mid)
Post: first tech job and promotions; increased academic readiness — Rosura

Synthesis: Multi-factor drivers of confidence (from this dataset)

  • Stage progression itself is associated with improved confidence (Pre = universally low; Mid = diversified with 33% confident/very and 50% somewhat).
  • Building projects helps, but quality/feedback matters more than mere completion (very high confidence appears with multiple apps plus mentor validation).
  • Problem-solving practice and peer belonging can raise confidence without a completed app (challenge-based learning, community).
  • Test scores alone do not predict confidence (near-zero correlation in Mid; high scorers can still feel low confidence if debugging is hard).
  • Debugging proficiency is a pivotal bottleneck; targeted support likely increases confidence for otherwise strong learners.
  • Consistency of participation reduces overwhelm and supports confidence growth.

Program moves to maximize confidence (data-grounded)

  • Make mentor feedback loops explicit and frequent around shipped mini-projects (mirror Rosura’s experience).
  • Add structured debugging clinics and “why it broke” walkthroughs to convert high-score/low-confidence profiles (e.g., Ying).
  • Guarantee regular challenge sets and peer-pair rotations to replicate Triana’s confidence gains without requiring full apps.
  • Implement attendance nudges and catch-up tracks to reduce overwhelm for intermittent participants (e.g., Pravalika).
  • Stage-tailored milestones: quick early wins at Pre; iterative projects + mentor reviews at Mid; showcase portfolios Post.

Methods and limitations

  • Confidence coding: Not=1, Somewhat=2, Confident=3, Very=4.
  • Correlation computed on Mid-stage (n=6) between test score and confidence; result ~0.00.
  • Small sample with missing data in several fields; interpret trends directionally.
  • Resume workshop attendance is universal at Mid, so its isolated effect on confidence cannot be determined from this dataset.

Appendix: Mid-stage snapshot

Name Score Web App Confidence Notes
Rosura 68 Yes Very confident Multiple apps; mentor praise
Enas 70 Yes Somewhat Basic code OK; complex is hard
Irina 44 Yes Somewhat Basic code OK; complex is hard
Ying 92 Yes Not confident Debugging issues dampen confidence
Triana 85 No Confident Challenge practice; peer support
Pravalika 28 No Somewhat Inconsistent attendance; overwhelmed