A hands-on map of the quantitative strand.

This reference explains what the closed-ended survey items and the national datasets contribute to a qualitative-dominant convergent mixed methods study of pedagogical friction, and what they are not built to do. Adjust the assumptions, then use the plain-language scripts to explain the plan without turning the study into a statistics lesson.

The big picture in one minute

The quantitative analysis is not trying to prove the dissertation through numbers. The proposal commits it to four bounded jobs: descriptive statistics, role-based comparisons, cross-tabulations where cell sizes permit, and domain-level summaries aligned to the framework. Everything else is out of scope by design.

For a methods explanation

"This is a qualitative-dominant convergent mixed methods design. The survey describes patterns across a wider group of educators, while the interviews, card sorts, and documents carry the interpretive work. The strands are integrated through joint displays rather than read as separate results."

For a practitioner audience

"I am using the survey to look for patterns in how educators experience AI-related learning conditions, not to rank schools or make sweeping claims about all educators."

For a committee question

"The quantitative strand is intentionally bounded. It supports descriptive patterns, role-based comparisons, and domain-level summaries. It is not designed to validate the framework as a psychometric scale, and it will not be reported as if it had."

Simulation controls

Use generated demonstration data to test what the measurements mean before applying them to real survey exports.

Sample size 200

Below an achieved shared-core sample of 80, analysis is limited to descriptive frequencies and role-based patterns.

Items per scale 6

Reliability estimates become more stable when item sets are coherent and sufficiently structured.

Item coherence 72%

Higher coherence means items move together as a scale.

Missing / "Don't know" 8%

For infrastructural items, "Don't know" can be a substantive finding and missing for scale scoring.

Achieved sample 200

At or above 80: descriptives plus role comparisons and domain summaries.

Inter-item coherence .72

Preliminary evidence of how items behave in this sample only.

Domain mean 3.62

Mean scoring keeps the five-point metric interpretable.

Usable for comparison Yes

Branching and missingness leave enough per-item data to compare roles.

How the items behave in this sample

Simulated item-total and inter-item correlations. The proposal commits to reporting these only as preliminary evidence of how items behave within this sample — never as validation of the framework as a scale.

How to read it

Item-total and inter-item correlations describe whether related items moved together here.
They are descriptive of this sample. They do not establish that the construct is measured well.
Early item statistics inform interpretation and future refinement, not mid-collection validation.

Say it simply

"I am checking whether questions meant to capture the same idea actually moved together for the people who answered. That is a description of this sample, not a claim that I have validated a scale."

Domain-level summaries on the original response metric

Item response pattern

Why mean scoring

Mean scoring keeps domain summaries on the original one-to-five response metric, so 3.6 stays interpretable as a position respondents actually chose rather than an arbitrary sum.

The three learner-facing domains — noetic, rhetorical, existential — are summarized together. Infrastructural items are reported separately because the framework treats infrastructural friction as the conditioning layer that enables or constrains the other three, not as a fourth peer domain.

Say it simply

"Instead of reading every item alone, I average related items into a domain summary. That lets me describe something like noetic friction while keeping the number on the same one-to-five scale people answered. Infrastructural items stay separate, because they describe the conditions around the other three rather than sitting beside them."

Role comparisons and small-cell cautions

Written AI policy

Small-cell risk

Effect size

Describe first, and often only describe

The plan is descriptive statistics, role-based comparisons, and cross-tabulations where cell sizes permit. Where role-based branching reduces the usable sample for particular items, the study reports descriptive evidence rather than unstable subgroup claims.

Recruitment uses a nonprobability frame. Because professional networks may overrepresent educators already engaged with AI, the achieved sample may overstate familiarity, policy awareness, and perceived support. Comparisons with national benchmarks are limited to variables with comparable definitions, and they do not make the sample representative or correct selection bias.

Say it simply

"I can compare roles when enough people in each role answered the item. When a group is too small, I describe the pattern instead of dressing it up as a test. And I say plainly that this sample probably leans toward educators already interested in AI."

Secondary datasets as context

Source	Useful for	Boundary
NCES School Pulse Panel	Published school-level weighted estimates of school conditions, AI-related practices, and policy signals.	Subgroup respondent counts are unavailable; poverty classifications use neighbourhood income-to-poverty ratios. Does not measure pedagogical friction.
RAND / Gallup American Educator Panel	Author calculations from public-use microdata weighted by `PORTAL_WEIGHT`, covering AI adoption and institutional supports.	Missing and role-inapplicable responses excluded, so n varies by item. Public files lack the complete survey design, so estimates are descriptive and no formal significance claims are made.

Bounded inference

These datasets situate participant accounts within broader K–12 patterns. They do not directly measure pedagogical friction and cannot validate the framework.

Because waves, item wording, populations, and equity grouping variables differ across sources, results are treated as convergent directional evidence — not change scores, and not a single comparable gap. Approximate standard errors use Kish’s weight-adjusted effective sample size. Aggregate outputs and reproducible scripts live in the public secondary-data companion (Miner, 2026a).

Say it simply

"NCES and RAND can show the national backdrop, such as AI policy or adoption patterns. They cannot directly tell me whether educators are preserving noetic, rhetorical, or existential friction."

Plain-language guide

Use this section when you need to explain the quantitative strand to a committee member, practitioner audience, or colleague who does not live inside measurement language.

1. What the survey is doing

The survey is looking for patterns in how educators report the conditions that preserve or erode meaningful learning work in the presence of generative AI.

2. What scales are doing

Related questions are grouped into domain summaries so the study can discuss noetic, rhetorical, and existential friction as interpretable learner-facing domains. Infrastructural items are summarized separately, as the conditioning layer around them.

3. What item statistics are doing

Item-total correlations, inter-item correlations, and internal consistency estimates describe how the items behaved for the people who answered. They are preliminary evidence about items, not a verdict on the theory.

4. What the 80-response floor is doing

If the achieved shared-core sample falls below 80, analysis is limited to descriptive frequencies, role-based patterns where appropriate, and narrative integration with the qualitative and structural evidence. It is a floor for describing, not a gate that unlocks stronger claims.

5. What group comparisons are doing

Comparisons may show whether reported practices differ by policy presence, role, grade band, or school context, but only when the sample and assumptions support that move.

6. What secondary data is doing

NCES and RAND provide national context. They help situate the study, but they are not direct measurements of pedagogical friction.

Companion comparison page

The Antigravity version provides a second interactive explanation of the same quantitative-methods ideas. Use it as a comparison view when you want a more analogy-driven walk-through of data cleaning, domain summaries, item behaviour, and the cautions around role comparisons.

Best use

Open it when you want a simpler teaching mode with analogies and guided examples that can support conversation with a less quantitative audience.

How it differs

This studio is the more polished research-reference version. The comparison page is more playful and tutorial-like, which makes it useful for checking whether the concepts are easy to explain.

Open comparison

Launch the companion quantitative measurements simulation.

Analysis flow

The workflow keeps early organization accessible while preserving a reproducible path for more advanced analysis.

Clean responses

Remove incomplete responses, check branch routing, recode Likert items, and flag "Don't know."

Describe patterns

Review frequencies, means, distributions, role groups, grade bands, experience, FRPL, urbanicity, and policy presence.

Summarize domains

Use mean scores for the three learner-facing domains, reporting infrastructural items separately as the conditioning layer.

Report item behaviour

Report item-total, inter-item, and internal consistency estimates as preliminary evidence about the items in this sample only.

Integrate carefully

Integrate through joint displays rather than reporting survey results on their own. The primary statistical record is kept in R, with selected headline estimates reproduced in jamovi as a verification check.

Interpretation rules

Report small subgroups descriptively rather than forcing inferential tests.
Treat a purposive, nonprobability survey as evidence of patterns within the study sample, not population estimates.
Keep "Don't know" visible on infrastructural items because uncertainty can be a finding about policy clarity and organizational conditions.
Keep the primary statistical record in R, reproducing selected headline estimates in jamovi as a verification check.
Never report item statistics as validation of the framework as a scale; they are preliminary evidence about items in this sample.
Integrate through joint displays: survey patterns describe, the qualitative strands interpret.

Decision point	Use when	Plain-language interpretation
Item-level descriptives	Always, especially for "Don't know" infrastructural responses.	Shows what respondents said before the data are compressed into scale scores.
Composite mean scores	Items are conceptually aligned and missingness is handled transparently.	Creates an interpretable dimension score that still lives on the one-to-five survey metric.
Item-total, inter-item, and internal consistency estimates	Each domain has enough items and response variation to compute them.	Describes how the items behaved in this sample. Explicitly not validation of the framework as a scale.
Role-based comparisons and cross-tabulations	Cell sizes permit, and branching has not thinned the item beyond use.	Describes how responses differ across roles. Reported descriptively where cells are small.
Domain-level summaries	Achieved shared-core sample is at or above 80.	Summarizes the learner-facing domains, with infrastructural items reported separately.
Descriptive frequencies only	Achieved shared-core sample falls below 80.	Frequencies, role-based patterns where appropriate, and narrative integration with the qualitative evidence.

Quick glossary

Short definitions for the terms most likely to need explanation in conversation.

Likert item

A single survey question with ordered response choices, often from strongly disagree to strongly agree.

Composite score

An average of related items used to represent a broader concept, such as infrastructural friction.

Internal consistency

A check of whether items meant to measure the same idea tend to move together.

Item-total correlation

How closely a single item tracks the summary of its domain. Descriptive of this sample.

Nonprobability frame

Recruitment that does not give every member of the population a known chance of selection, so results describe respondents rather than estimating a population.

Convergent directional evidence

Agreement in direction across sources whose waves, wording, and populations differ. Not a change score and not a single comparable gap.

Joint display

A table or figure placing qualitative and quantitative evidence side by side so they are interpreted together rather than separately.

One-sentence anchor: The quantitative strand describes patterns that the qualitative strands then interpret.

Common caution: Coherent items do not confirm the framework. The survey is not designed to validate it as a psychometric scale.

Best framing: Bounded, transparent description rather than generalization or validation.

Privacy and use

Proposal stage. This page presents the planned quantitative strand. It reports no findings, and no participant data have been collected. Every number shown in the simulation is generated demonstration data.

The page is static. It uses no external scripts, analytics, cookies, forms, or storage, and it holds no participant responses, consent records, or school or district identifiers. Any approved study data will be handled only through the IRB-approved secure workflow, using de-identified exports and an audit trail.