A 20% Response Rate, a Rewritten Hypothesis and a 19% Similarity Report
Priya, a doctoral candidate in Education, administers a 42-item questionnaire on digital-classroom readiness to 480 secondary teachers across one district. Only 96 usable responses return (20%). Among respondents, 71% teach in private aided urban schools, whereas the district teacher population is 55% rural government. Her registered hypothesis states that readiness does not differ by school type. After seeing the data, she rewrites it as 'rural government teachers are less ready' and reports p = .04 from an independent-samples t-test run after that rewrite. Her supervisor urges submission within three weeks to a Scopus-indexed journal whose limits are 10% overall similarity and 2% from any single source. The similarity report shows 19%; most matches are her own earlier conference paper (published with an ISBN) and the method paragraph of a 2019 article. Priya's fellowship ends in three weeks and she needs one publication to retain it.
1. Does Priya's p = .04 support her rewritten hypothesis? State precisely what she can and cannot claim, and what she should report instead.
Read the worked answer
No - the test does not support the rewritten hypothesis as a confirmatory result.
- The hypothesis was changed after inspecting the data, so it is post hoc. Testing a direction suggested by the same data and reporting the resulting p-value as if it had been predicted inflates Type I error; this is HARKing (hypothesising after the results are known).
- The one-tailed logic implied by 'less ready' is invalid here, because the direction came from the data, not from theory.
- The sample itself is compromised: a 20% response rate with urban private schools at 71% against a population share of 55% indicates non-response bias, so no district-level inference is safe.
What she may claim: in this convenience sample, rural government teachers scored lower; report group means, SDs, mean difference, Cohen's d and the 95% CI, and label the analysis explicitly as exploratory and hypothesis-generating.
What she should do: state in the paper that the registered hypothesis was different and that the comparison was post hoc; run and report a non-response check (compare early vs late responders, or weight to the known district profile); if a confirmatory claim is needed, pre-register and collect an independent sample, since splitting 96 cases gives roughly 48 per half and power near 0.5 for a moderate effect.
2. How should the 19% similarity report be handled before submission, and what is genuinely at stake in the three-week deadline?
Read the worked answer
Break the report down item by item rather than treating 19% as one number.
- Matches to her own conference paper. Reusing her own published text without attribution is text recycling; if the findings also overlap, it edges toward duplicate publication. She must paraphrase the overlapping passages, cite the conference paper, and disclose it in the cover letter to the editor so the overlap is transparent.
- The 2019 method paragraph. Method boilerplate is still someone's prose. She should rewrite it in her own words and cite the source for the procedure (not for the wording). Running it through a paraphrasing tool to defeat detection is misconduct, not a fix.
- Limits. At 19% overall she is already above the journal's 10% ceiling and can expect a desk rejection or a misconduct query that follows the paper.
What is at stake in three weeks: rewriting methods, re-running the analysis with the non-response check, and obtaining supervisor and co-author sign-off is realistically two to three weeks of work with no margin. The defensible move is to request an extension or target a later issue rather than submit a report certain to be flagged. She should also keep the similarity report and her revision log as institutional records, since fellowship-linked publications are often audited later.