An APA methods section for scale development explains how the construct became a set of items, how participants were recruited, and how the measurement model was tested. The CFA fit statistics belong with the results. Keeping the analysis plan separate from its findings makes both easier to evaluate.
Below are a reporting outline, worked Methods and Results paragraphs, and a completed CFA fit table. The examples use explicitly fictional values. APA's quantitative reporting standards, including the structural equation modeling guidance, provide the broader reporting framework (Appelbaum et al., 2018).
What Belongs in Methods and What Belongs in Results?
The Methods section explains what you planned and did. The Results section presents what those procedures produced. An item deletion rule belongs in Methods; the items actually removed and the consequences belong in Results. If a decision followed inspection of the results, identify it as such instead of presenting it as an advance plan.
| Reporting question | Methods | Results |
|---|---|---|
| How was the scale developed? | Construct definition, item sources, expert review, and pilot procedures | Findings from each development stage and resulting revisions |
| Who supplied the data? | Recruitment, eligibility, collection dates, consent, planned and achieved sample sizes, and exclusion and missing-data procedures | Participant flow, exclusions applied, and missing-data diagnostics |
| How was the factor structure investigated? | EFA/CFA samples, estimator, model specification, and evaluation strategy | Factor solution, fit statistics, parameter estimates, and diagnostic findings |
| Was the model changed? | Planned revision rules, if any | Actual changes, their rationale, and which analyses were exploratory |
The table offers one way to organize the report; placement can vary with a journal's structure. Methods can describe the achieved sample as well as the sampling plan. Distinguish decisions made before examining the evidence from changes prompted by it. The SEM reporting recommendations explicitly ask authors to disclose when their models were specified and how revisions were justified (Hoyle & Isherwood, 2013).
A Practical Outline for a Scale Development Methods Section
Describe the construct and the process used to write and review items. Record who reviewed the content, what they evaluated, and how their judgments informed revisions. Name any content validity index you actually calculated and justify its interpretation for your panel and procedure.
Document recruitment and eligibility for each sample. Explain consent and ethics review where applicable, the collection period, sample size planning, and missing data procedures. Make the relationship between exploratory and confirmatory samples explicit. Report repeated use of the same participants rather than describing it as independent confirmation.
The analysis subsection should identify the software and version, estimator, treatment of item responses, model specification, and basis for evaluating fit. Report whether the analysis used participant responses or simulated data (Appelbaum et al., 2018). For the earlier decisions, see the EFA analysis guide and CFA fit index guide.
Worked Methods Paragraph: Describing a CFA
Fictional reporting example. All study details and numbers below are illustrative, not participant findings or modidoc output. Replace them with your own verified study details and results.
The scenario assumes a previously specified scale with 12 continuous indicators and three correlated factors. Each factor has four indicators. The example uses ordinary maximum likelihood with the sample covariance based on N − 1. Ordinal responses or robust estimation require a description and output appropriate to those choices.
We evaluated the proposed three-factor structure in a separate confirmation sample of 300 respondents. The 12 items used 0–100 visual analogue response scales and entered the analysis as continuous indicators. We fitted the covariance structure by maximum likelihood. Each item loaded on its designated factor, one loading per factor was fixed to 1 for identification, and factor covariances were freely estimated. Cross-loadings and residual covariances were fixed to zero. We examined convergence and admissibility before interpreting fit. Evaluation included the chi-square test, comparative fit index (CFI), Tucker–Lewis index (TLI), root mean square error of approximation (RMSEA) with its 90% confidence interval, standardized root mean square residual (SRMR), and inspection of individual residuals. We planned no modifications to the confirmation model.
This is an analysis paragraph, not an entire Methods section. A submission still needs the study-specific recruitment and sample size rationale, software version, evidence supporting the estimator, missing data procedures, and sources supporting its interpretation strategy. Supply those details from your records.
How to Report CFA Results in APA Format
Name the model, report its test statistic and degrees of freedom, and identify the estimator. Present the selected fit indices with the RMSEA confidence interval. Parameter estimates and local fit also need attention; a fit table alone does not document a measurement model (Hoyle & Isherwood, 2013).
Fictional Results paragraph, continuing the same example:
The prespecified three-factor model converged without inadmissible estimates. The test of exact fit yielded χ²(51) = 78.00, p = .009. Approximate fit estimates were CFI = .955, TLI = 0.942, RMSEA = 0.042, 90% CI [0.022, 0.060], and SRMR = 0.045. No items were deleted, and no residual covariances were added. The chi-square test rejected exact fit; the approximate indices should therefore be considered alongside the residuals and parameter estimates rather than treated as proof of a valid scale.
Table 1
Global Fit of the Fictional Three-Factor CFA Model
| Model | N | χ² | df | p | CFI | TLI | RMSEA | 90% CI for RMSEA | SRMR |
|---|---|---|---|---|---|---|---|---|---|
| Three correlated factors, 12 indicators | 300 | 78.00 | 51 | .009 | .955 | 0.942 | 0.042 | [0.022, 0.060] | 0.045 |
Note. All values are fictional. CFI = comparative fit index; TLI = Tucker–Lewis index; RMSEA = root mean square error of approximation; CI = confidence interval; SRMR = standardized root mean square residual. The illustration uses ordinary maximum likelihood, not robust or scaled statistics.
The paragraph and table demonstrate two presentation options. In a manuscript using the table, shorten the paragraph to the findings needed for interpretation instead of repeating every cell. Add a separate parameter table with the estimated loadings and other model parameters, their uncertainty, and clear labels distinguishing standardized from unstandardized estimates (Appelbaum et al., 2018).
APA Formatting for CFA Statistics
APA's Numbers and Statistics Guide distinguishes Latin statistical symbols from Greek letters. Italicize N, df, and p. Keep the Greek χ in χ² in standard type. Define fit index abbreviations at first use.
Report exact p values to two or three decimal places, except values below .001, which are reported as p < .001. A displayed software value of 0.000 is not a probability of zero. Omit the leading zero for p values and correlations.
Rounding should preserve information needed to interpret the result. This example uses three decimals for fit indices so nearby values remain distinguishable; that is an editorial choice, not an APA requirement that every CFA index use a fixed number of decimals. Follow the target journal's instructions consistently.
Disclosing AI Assistance
Check the target journal's current policy. APA Journals requires disclosure and attribution of relevant generative AI use, holds authors responsible for verifying content, and excludes AI tools from authorship. Its guidance varies the placement of disclosure with the activity performed, so one declaration location will not suit every use (APA Journals policy).
A disclosure for item generation should name the tool and model or version, describe the task, identify what entered the item pool, and explain the human review. Preserve the prompts and outputs needed to substantiate that account. Language editing, analysis code generation, and item generation are different activities and should be described accurately.
Using modidoc for a Methods Draft
modidoc's methods drafting stage helps organize the development record into text for researcher review. Keep simulation-based screening separate from validation using participant responses. If empirical analysis has not been completed, the draft should identify that work as pending. Before submission, match every reported statistic to its analysis output and replace any placeholders with verified study details.
Frequently Asked Questions
How do I report a scale development process in the methods section?
Explain how you defined the construct, developed and reviewed items, recruited participants, and planned the analysis. Identify the data and decisions used at each stage so readers can follow the development record. Use the applicable JARS checklist when preparing the final manuscript (Appelbaum et al., 2018).
How do I report CFA results in APA format?
Report the model, estimator, χ² with degrees of freedom and p, relevant fit indices, and RMSEA confidence interval. Include parameter estimates and diagnostic findings elsewhere in the results. The worked example above shows the format; its values are fictional.
How do I disclose the use of AI tools in a manuscript?
Follow the journal's current instructions and describe the tool, purpose, outputs used, and human verification. Disclose substantive research uses separately from editing. Do not claim that an expert review or statistical check occurred unless it actually did.
References
American Psychological Association. (2022). APA Style numbers and statistics guide. 2022 edition.
American Psychological Association. (2025, August). APA Journals policy on generative AI: Additional guidance. Publisher policy.
Appelbaum, M., Cooper, H., Kline, R. B., Mayo-Wilson, E., Nezu, A. M., & Rao, S. M. (2018). Journal article reporting standards for quantitative research in psychology: The APA Publications and Communications Board task force report. American Psychologist, 73(1), 3–25. https://doi.org/10.1037/amp0000191.
Hoyle, R. H., & Isherwood, J. C. (2013). Reporting results from structural equation modeling analyses in Archives of Scientific Psychology. Archives of Scientific Psychology, 1(1), 14–22. https://doi.org/10.1037/arc0000004.
Previous: Final Checks Before Fielding a Questionnaire
First in the series: Construct Extraction and Factor Structuring