Free scoping read. A written assessment of your manuscript in two business days, at no cost.

Start one
Contents

Academy · Research Data, Analytics & AI · Biostatistics Self-Learning

Biostatistics as a Self-Learning Experiment in the Health Sciences: Evidence From Comparative Teaching Studies

Several decades of comparative teaching research suggest that health science students often learn biostatistics more effectively when the course is organised around self-paced, problem-centred work rather than passive lectures.

Pubrica AcademyWritten by a biostatistics education researcher Reviewed bya senior biostatistician Published10 October 2026 Last checked10 October 2026 Reading time7 minutes Sources10 peer-reviewed

Biostatistics has a reputation, whether deserved or not, as the course that medical and health science students find most daunting. Decades before e-learning platforms existed, educators were already testing whether students learn the subject more effectively through lectures or through self-paced, problem-based material. The question has implications beyond pedagogy: a clinician or researcher who does not develop a working intuition for statistical reasoning during training is unlikely to develop it later, under time pressure, while interpreting another investigator’s trial.

This article examines what comparative teaching studies found when self-instructional or problem-based biostatistics courses were compared directly with traditional lecture formats, and what that evidence implies for how research-support teams should train investigators who need to build statistical competence quickly.

Learning objectives

  • What the earliest controlled comparisons of self-paced and lecture-based biostatistics teaching measured
  • Why self-selection into self-directed formats complicates the interpretation of improved outcomes
  • How more recent problem-based learning (PBL) trials in biostatistics compare to the older independent-study evidence
  • Practical implications for structuring in-house biostatistics training for clinical investigators

Common misconceptions

  • “Self-directed learning is just lecture material minus the lecturer.” In the studies examined here, self-instructional and problem-based formats changed the whole structure of the course — self-paced modules, case-based problem sets, and student-controlled timing — not merely the removal of a live instructor from otherwise identical content.
  • “If students taught themselves scored higher, the teaching method caused the gain.” In the independent-study programme described by Hebel and Morton, students who opted into the self-paced track already had higher MCAT scores and stronger first-year grades than those who stayed in lectures, so self-selection — not the format alone — likely explains part of the exam gap.
  • “Biostatistics is equally hard to self-teach regardless of course design.” The problem-based learning trial in Indian medical students found a statistically significant score advantage (P < 0.05) for the PBL group over the didactic-lecture group across objective, descriptive, and viva-voce components, suggesting course design materially changes outcomes, not just student effort.

Origins of self-instruction in biostatistics teaching

Long before “flipped classrooms” became a widely used term, medical schools were experimenting with self-instructional biostatistics manuals and self-paced independent study tracks as alternatives to standard lectures1. The rationale was that biostatistics is a skills-intensive subject, better learned by working through problems at the learner’s own pace than by passively absorbing a lecture delivered at a pace set for the whole class.

Outcomes measured in the independent-study programme

One of the earliest controlled comparisons arose from a combined epidemiology and biostatistics course for second-year medical students, where an Independent Study Programme (ISP) was offered as an opt-in alternative to the standard lecture track1. Forty of 176 students chose the self-paced ISP. On 35 identical final-exam questions shared with the lecture group, ISP students scored higher, and the advantage persisted after statistically adjusting for MCAT entrance scores1. Students who completed the ISP reported high satisfaction and requested that the option be extended to later courses1.

Notably, the time a student took to complete the self-paced modules did not correlate with exam performance1 — pacing itself was therefore not the active ingredient; some aspect of direct engagement with the material, rather than passive reception of it, appears to have been.

Recent comparisons of problem-based learning

A more recent randomized comparison at an Indian medical college split 96 final-year students into a didactic-lecture group and a problem-based learning (PBL) group taught the same content by the same instructors2. Assessed two weeks later across objective, descriptive, and viva-voce components, the PBL group scored significantly higher than the lecture group (P < 0.05)2. The authors attributed the gap to PBL's emphasis on analytical reasoning and self-directed problem-solving rather than memorized procedure2.

Self-selection as a limitation of the evidence

The independent-study result carries an important caveat: students were not randomly assigned to the self-paced track but chose it, and the self-selected group was already academically stronger on entry1. Consequently, some, though not necessarily all, of the exam gap probably reflects who opted in rather than what the format taught. The more recent PBL trial addresses this limitation more rigorously by assigning students to groups and controlling instructor and content, which is why its significant result constitutes stronger evidence for format effects specifically2.

Read together, the two studies, separated by four decades, point in the same direction without being identical experiments: active, self-paced or problem-centred engagement with biostatistics content tends to outperform passive lecture delivery, although the oldest evidence may overstate the effect somewhat because of self-selection.

Implications for training clinical investigators

For teams supporting clinical investigators who need working statistical literacy quickly, rather than a semester-long course, the implication is to favour case-based, problem-first materials, built around the investigator’s own dataset or protocol, over generic slide decks. Allowing researchers to work through a real sample-size calculation or randomization scheme for their own study, with support available, replicates the active-engagement conditions under which both studies found an advantage.

The broader evidence base on active learning

The biostatistics-specific findings above sit within a much larger and well-replicated evidence base on active versus passive instruction across science, technology, engineering and mathematics (STEM) subjects generally. A landmark 2014 meta-analysis in PNAS pooled 225 studies comparing active learning with traditional lecturing across undergraduate science, engineering, and mathematics courses and found that average examination and concept-inventory scores rose by about 6% (a standardized effect of 0.47 standard deviations) under active learning, while failure rates fell from 33.8% under straight lecturing to 21.8% under active learning — equivalent to students in lecture-based courses failing at roughly 1.5 times the rate of those in active-learning courses3. Evidence on this scale, spanning 225 independent studies, is part of the reason why the discipline-specific biostatistics findings described above are plausible rather than an artefact of two small, idiosyncratic trials: they align with one of the most robustly replicated findings in science education research.

Variation in the effectiveness of active learning formats

It would be a mistake to read the evidence as indicating that any active format outperforms any lecture. Two more recent and more cautious analyses complicate the picture considerably. A 2017 systematic review of 11 studies specifically testing flipped-classroom formats in medical education found the results “equivocal”: four studies showed a significant score improvement, four showed no difference and three were mixed, with the only consistent signal being a possible benefit on higher-order cognitive tasks rather than on exam scores overall4. More notably, a 2022 meta-analysis pooling seven controlled studies (622 participants) testing problem-based learning specifically in first-year medical students found no significant difference from conventional teaching on critical thinking, problem-solving, or self-directed learning, with very high heterogeneity across the included studies10. Taken together with the stronger positive findings described earlier, the most accurate summary is that active and self-directed formats have a real, replicated average advantage across STEM broadly, but that this advantage is neither universal nor guaranteed by labelling a course “active” or “self-paced”; implementation quality, subject matter and learner population all appear to moderate the effect substantially.

Statistical misconceptions among clinicians

The need to improve biostatistics teaching is a practical rather than a hypothetical concern. A 2021 online survey of 376 critical care physicians and respiratory therapists in Argentina found that 63.1% did not demonstrate adequate understanding of the p-value, and only 27.6% could correctly identify its formal definition; just 7.4% passed both a theoretical and a practical test of p-value interpretation5. Respondents without any formal training in research methodology were more than twice as likely to fail the knowledge check (adjusted odds ratio 2.50), and those who read fewer than six scientific articles a year were over three times as likely to fail (adjusted odds ratio 3.27)5. These figures describe practising clinicians making real treatment decisions, not students facing an examination, which is why the format of biostatistics training, and not merely its existence, warrants this level of scrutiny.

Spaced, blended and remote delivery

Two further strands of evidence bear directly on how, not just whether, to structure self-directed statistics training. First, a classic meta-analysis of distributed-practice research found that spacing study sessions apart, rather than massing them together, reliably improves long-term retention across verbal recall tasks — a finding directly applicable to structuring a multi-week biostatistics module around spaced problem sets rather than a single intensive workshop6. Second, more recent, direct tests of delivery format specifically in biostatistics have found blended and fully online formats performing as well as, or better than, face-to-face teaching: a randomized comparison of online versus on-site training in health research methodology among Indian scientists found similar knowledge gains in both formats7, and a 2022 controlled comparison of blended versus face-to-face applied biostatistics teaching for postgraduate medical students found the blended group scored significantly higher on the final exam (mean 16.69 vs. 14.49, P = 0.013)8. A separate strand of research on statistics anxiety is also relevant here: a frequently cited review found statistics anxiety is common among university students generally and is associated with avoidance behaviors and poorer performance, which strengthens the case for self-paced, lower-stakes formats that let a learner work through difficult material without the time pressure and public exposure of a live lecture setting9.

Key points

Key points

Across two very different eras of teaching research, biostatistics taught through self-paced, problem-centred formats outperformed passive lecture delivery, though the oldest evidence is confounded by self-selection.

Practical application

Practical application

  1. Prefer problem-based materials to slide-based lectures. In-house biostatistics training should be built around a real protocol, dataset or decision that the investigator must address, rather than generic worked examples.
  2. Separate pacing from active engagement when evaluating a course. The independent-study data indicate that time to completion did not predict performance; a self-paced format alone should therefore not be assumed to help unless it is also problem-centred.
  3. Exercise caution in generalising from opt-in comparisons. Where staff self-select into a new training format, part of any measured improvement may reflect who chose the format rather than what it taught.

Frequently asked questions

The comparative evidence leans toward self-paced or problem-based formats producing better exam performance than straight lectures, but the strongest supporting study is a small, single-institution randomized comparison, not a large multi-site trial.

No. In the independent-study programme, completion time did not correlate with exam performance, suggesting that content structure and active engagement mattered more than speed.

The underlying principle — active, problem-centred engagement outperforming passive delivery — is domain-general, but neither study tested postgraduate researchers directly, so the extrapolation should be treated as reasonable, not proven.

References

  1. Hebel JR, Morton RF. Teaching experience with an independent study programme in epidemiology and biostatistics. International Journal of Epidemiology. 1979;8(2):177-180. read 10 October 2026. View source
  2. Effectiveness of problem-based learning approach for teaching-learning biostatistics among medical students. Journal of Education and Health Promotion. read 10 October 2026. View source
  3. Freeman S, Eddy SL, McDonough M, Smith MK, Okoroafor N, Jordt H, Wenderoth MP. Active learning increases student performance in science, engineering, and mathematics. Proceedings of the National Academy of Sciences. 2014;111(23):8410-8415. read 10 October 2026. View source
  4. Hughes Y, Lyons N, et al. Does the Flipped Classroom improve exam performance in medical education? A systematic review. MedEdPublish. 2017. read 10 October 2026. View source
  5. Andreu MF, Diaz Ballve LP, Verdecchia DH, Monzón AM, Dias de Carvalho T. Is the p-value properly interpreted by critical care professionals? Online survey. Revista Brasileira de Terapia Intensiva. 2021;33(1):88-95. read 10 October 2026. View source
  6. Cepeda NJ, Pashler H, Vul E, Wixted JT, Rohrer D. Distributed practice in verbal recall tasks: a review and quantitative synthesis. Psychological Bulletin. 2006;132(3):354-380. read 10 October 2026. View on PubMed
  7. Kanchanaraksa S, Sisson SD, Aggarwal A, et al. A comparison of online versus on-site training in health research methodology: a randomized study. BMC Medical Education. 2011;11(1):37. read 10 October 2026. View source
  8. Shahsavari S, Jambarsang S. The effect of blended learning on teaching applied biostatistics for postgraduate medical students. Journal of Medical Education and Development. 2022;17(1). read 10 October 2026. View source
  9. Onwuegbuzie AJ, Wilson VA. Statistics anxiety: nature, etiology, antecedents, effects, and treatments--a comprehensive review of the literature. Teaching in Higher Education. 2003;8(2):195-209. read 10 October 2026. View source
  10. Manuaba IBAP, Yi-No, Wu CC. The effectiveness of problem based learning in improving critical thinking, problem-solving and self-directed learning in first-year medical students: a meta-analysis. PLoS ONE. 2022;17(11):e0277339. read 10 October 2026. View source

Every source above is a peer-reviewed article, a publisher’s record of one, or a standard issued by the body that owns it. Each carries the date it was last read. Where a standard moves, this article is revised rather than left up with a note.

How to cite this article

Pick your referencing style and copy it. Your university or target journal may vary the punctuation — check its guide if it has one.

Pubrica Academy. (2026, October 10). Biostatistics as a Self-Learning Experiment in the Health Sciences: Evidence From Comparative Teaching Studies. Pubrica. https://pubrica.com/academy/concepts-definitions/biostatistics-an-experiment-with-self-learning-in-the-health-sciences/

This article is written to be learned from. Cite it if you are referring to what it says; if you are using the evidence behind it, cite the sources listed above instead.

Self-paced biostatistics training checklist

Use this when building in-house statistics training for clinical investigators.

[ ] Anchor the course to a real protocol or dataset, not generic examples
[ ] Structure content as problems to solve, not slides to review
[ ] Allow self-paced progression but track engagement, not just time spent
[ ] Assess with applied tasks (e.g., a real sample-size calculation), not recall questions
[ ] Flag confounds if adoption is opt-in rather than assigned

Pubrica AcademyWritten and maintained by Pubrica’s subject matter experts. Articles are revised when the underlying standards change, and each article states the date on which it was last reviewed.

How Pubrica can help

Pubrica’s biostatisticians support investigators through statistical problems drawn from their own protocols rather than generic instruction, reflecting the active-engagement conditions that the evidence favors.

Find something else

Search the Pubrica Academy

Access is free and requires no account. Enter a topic, or open the full library to filter by category, sub-topic or resource type.

Development of statistical competence through study-based problems

A scoping assessment is free and takes two business days. It provides a written assessment of what the work requires and which specialist would undertake it.

Get a free quote

Regulations

Check your project against this rule

  1. 1Your work
  2. 2Your details
  3. 3Send
What do you need?

Material you send is held under signed confidentiality, with access limited to the assigned team and encrypted transfer. It is not used to train anything and is not passed to third-party tools. Ask for the agreement before you send anything you would rather cover first.