# MethodSpring instructor guide

MethodSpring is a free learning companion for research methods, measurement, and statistics. It supports individual study and export-based classroom use. Its instructor showcase demonstrates future class administration with synthetic data.

## Capability status

| Capability | Status and classroom use |
| --- | --- |
| Lessons, practice, labs, walkthroughs, reporting resources | Available: share activity links through your course system. |
| Blueprint & Coverage and QA Validator | Available: inspect the current bank and export coverage. Counts are operational checks, not proof of teaching quality. |
| Individual progress and exports | Available: students sign in to sync or export a guest backup. |
| Cohort dashboard and assignment builder | Demonstration: all people/results are synthetic; previews are not delivered. |
| AI formative feedback | Disabled until quality evaluation passes and the service is configured. |
| Live rosters, assignment delivery, gradebook and LMS synchronization | Planned; do not represent them as available. |

## A classroom sequence

1. Select learning outcomes and inspect their prerequisite lessons.
2. Ask students to complete a topic diagnostic. Explain its limited coverage.
3. Share a lesson, worked case, and output interpretation lab.
4. Ask for independent practice and a results paragraph.
5. Review student-provided exports and distinguish objective, assisted, and self-assessed work.
6. Use a contrasting case to address recurring misconceptions.
7. Complete a post-test, then a delayed check seven or more days later.

The interactive showcase illustrates this sequence. It operates entirely in the page and has a Reset demonstration button. It does not alter the instructor's or a student's learning profile.

## Annotated 50-minute lesson: one-way ANOVA

| Time | Activity | Teaching purpose |
| --- | --- | --- |
| 0–5 min | Present three independent groups and a numerical outcome; state the question. | Establish design fit before a test name. |
| 5–13 min | Ask for variables, hypotheses, and prerequisite concepts. | Surface misconceptions about omnibus conclusions. |
| 13–25 min | Read the guided walkthrough's supplied output. | Connect assumptions, F, df, p, effect size, and follow-ups. |
| 25–40 min | Paired output interpretation followed by an independent practice response. | Require evidence and explanation, not a p-value alone. |
| 40–47 min | Draft a results paragraph and compare with the rubric. | Practise responsible reporting. |
| 47–50 min | Exit ticket: one justified conclusion and one limitation. | Identify the next teaching step. |

## Six-week supplement

1. Research questions, variables, sampling, uncertainty, and design selection.
2. One-way ANOVA, assumptions, effect sizes, and reporting.
3. Repeated measures, factorial/mixed designs, interactions, and ANCOVA.
4. Regression, diagnostics, MANOVA, and classification decisions.
5. Measurement reliability/validity and introductory factor analysis.
6. Synthesis, independent application, post-test, and later retention check.

Advanced IRT, SEM, and HLM can replace or extend weeks 5–6 for prepared cohorts. Use the prerequisite bridge rather than assuming algebra or measurement knowledge.

## Semester curriculum map

| Weeks | Focus | Example evidence of learning |
| --- | --- | --- |
| 1–3 | Foundations and research reasoning | Explain a design, identify confounding, interpret uncertainty. |
| 4–6 | ANOVA, repeated/factorial designs, ANCOVA | Justify assumptions and follow-ups in a complete case. |
| 7–8 | Regression, MANOVA, classification | Interpret model/output choices in context. |
| 9–11 | Measurement, EFA/CFA, IRT | Connect score interpretation, model assumptions, and diagnostics. |
| 12–14 | SEM and multilevel/growth models | Explain identification, dependence, centering, fit, and limits. |
| 15 | Integration and delayed review | Produce an independent methods/results rationale. |

## Rubric example: supported results paragraph

- Design and variables are named accurately: 2 points.
- Appropriate statistic, degrees of freedom, and p-value are reported: 3 points.
- Effect size and uncertainty are interpreted in context: 2 points.
- Assumptions, follow-ups, and limitations are justified: 3 points.

Adapt to the method and learning outcomes. A model's formative score must not determine a formal grade. Explain rubric criteria before collecting work.

## Current export-based workflow

Share the student's export instructions and define collection checkpoints: after the diagnostic, after a study path, and before an exam. Collect exports through your existing approved course system. A JSON file contains individual events; CSV reports can be used for coverage or demonstration data. Never import students' histories into an instructor's personal learning account to aggregate a class. Review or aggregate copies in an authorized separate environment, with the appropriate consent and retention policy.

Interpret a progress export cautiously: self-assessed responses, assisted practice, and reading completion are different measures. Instructional answer keys are available in the browser; this is a study environment, not a secure examination platform. Repeated or exposed items cannot establish fresh transfer.

## Content quality and release review

Verify dataset/model alignment and numerical fixtures. Review explanations, distractors, ambiguous prompts, prerequisites, and reporting claims. The old archive QA reports only establish historical count gates. Qualified reviewers should record a name, date, version, and disposition before a lesson is labeled reviewed.

Some walkthroughs use illustrative numerical cases separate from downloadable datasets. Match data and model before claiming reproduction. Optional R scripts include descriptive checks, package version reporting, and explicit model assumptions.

AI release evaluation requires at least 100 reviewed cases across all families, at least 95% acceptable explanations, zero critical statistical errors, and rubric scores within one point of reviewer scores for at least 90% of applicable cases. Failed workflows stay disabled.