Work

Preregistration: Forced answers from non-eligible respondents — required items, ghost responses and late response in three SMS-invited web surveys of Korean dentists

Year: 2026
Type: preprint
Abstract: This is a preregistration of a secondary-data analysis, deposited before any confirmatory model was run. It follows the Open Science Framework / Center for Open Science template for preregistration of secondary data analysis (Van den Akker et al., 2021). Question. When a required web-survey item has no exit option, do respondents who are not eligible for it still give a substantive answer ("ghost answers")? And do respondents who answer late differ from early respondents in job satisfaction and in response quality? Data. Three existing web surveys of Korean dentists, each sent as a census-type invitation by Korean Dental Association member text message and administered with Google Forms: 2019 (n = 2,363; two questionnaire versions analysed as one sample, with version as a covariate), 2024 (n = 2,059 before data cleaning; 2,033 after) and 2026 (n = 495 before cleaning; 463 after). No send log exists, so the first recorded submission in each wave stands in for the send time. Quality analyses use the files before cleaning, because cleaning had already removed respondents who straight-lined all professional-values items. Hypotheses. H2 (ghost answers): H2a and H2b are descriptive, giving the share of non-eligible respondents who answered by item design (required without exit option, required numeric, optional, skip-logic). H2c is confirmatory: among 2026 respondents with no relocation plan who faced the same required reason item, those who chose a listed reason (n = 366) are compared with those who wrote "not applicable" (n = 49) on a composite of selecting too many options and work-time shares not summing to 100 (logistic regression, risk difference with Newcombe interval). H1: late responders (last 25% of submissions within each survey) and job satisfaction, as a standardised difference pooled across surveys with an equivalence test (bounds ±0.20). H3: late responders and any response-quality problem, two-sided, as an unadjusted pooled odds ratio, with the adjusted odds ratio reported alongside. Composition of late versus early responders is shown descriptively. Contents. Data sources and cleaning rules, a disclosure of everything the authors had already inspected (including submission-time distributions, reason distributions and straight-lining by ghost status; no late-response outcome table or model had been produced), variable definitions, samples, models, multiplicity control (Holm for three confirmatory tests; Benjamini–Hochberg for exploratory results), prespecified sensitivity analyses, independent recoding of all indicators by two co-authors before analysis, a publication rule and a deviation policy. Ethics: Institutional Review Board of Seoul National University School of Dentistry (2019 survey 20190015; 2024 survey S-D20240015; 2026 survey S-D20250017). Individual-level data cannot be shared publicly under the IRB approvals; analysis code will be deposited at publication. Funding: Institute of Health Policy, Korean Dental Association; National Research Foundation of Korea (MSIT) grant RS-2026-25593489.
Language: English
Open Access status: green
OpenAlex ID: W7236838609
Cites:
Cited by:
Related to: 10