ReReRe

A permutation-based individual-coherence method for detecting careless respondents in questionnaire data. Drop in a dataset; obtain flagged rows.

Runs entirely in your browser · no upload · no account

Your data never leaves this device — how we guarantee it →

1 · Load your data

Drop a CSV or Excel file here — or click to browse
Rows = respondents, columns = questionnaire items (Likert-type). Accepts .csv / .tsv (comma, semicolon or tab) and .xlsx / .xls (you'll be asked which sheet to use). ID columns are detected and set aside.
Advanced settings
Computing…

2 · Results

–
respondents
–
items used
–
coupled pairs
–
flagged careless
Distribution of z-scores. Low z = careless (coherence below the respondent's own permutation baseline). Red line = current threshold; re-run after changing it in Advanced settings, or just re-download — flags are recomputed live from the threshold.
Columns added: z_rr, flagged, irv, longstring, person_total, n_missing

How it works

Respondents who answer attentively are internally consistent: on items that tap the same underlying construct they give compatible answers, so those items covary within the person just as they do across the sample. Careless respondents — answering at random, straight-lining, or drifting in and out of attention — break that internal covariance. ReReRe turns this into a per-respondent score by asking one question: did this person answer the questionnaire's strongly-related item pairs more coherently than they would by chance?

The method in four steps

  1. Find the coupled pairs. ReReRe correlates every pair of items across your sample and keeps the top fraction with the largest |r| (default 3%). These are the pairs that genuinely move together — typically items from the same scale or construct. They are the backbone the coherence check is built on.
  2. Score each person's coherence. For one respondent, it takes their answers to the two items in each coupled pair and measures how tightly those answers track across all the coupled pairs (an individual-level |correlation|). An attentive person tracks them closely; a careless one produces near-noise.
  3. Build a personal chance baseline. The same coherence is recomputed many times on random item pairs — a permutation done separately for each respondent. This reveals how coherent this particular person would look on unrelated items, absorbing their own response style, scale usage and quirks.
  4. Turn it into a z-score. The person's coupled-pair coherence is expressed in standard-deviation units above their own random baseline — z = (coupled − mean random) / sd random. A respondent is flagged when z ≤ threshold (default 1.5): their answers to items that should agree are no more coherent than their answers to items picked at random — the signature of careless responding.

Why the permutation baseline matters

Comparing each person to their own random baseline is what makes ReReRe self-calibrating. It assumes no multivariate normality, needs no external cut-off table, and requires no per-dataset tuning: the reference distribution is generated from the data itself, one respondent at a time. That is the crucial difference from distance-based outlier detectors such as Mahalanobis distance, whose chi-square thresholds assume a normality that Likert data violate and which, applied by the book, end up flagging almost no one.

What it detects — and what it doesn't

ReReRe targets inconsistent carelessness: random responding, partial or intermittent inattention, and any pattern that scrambles the cross-item structure. It is deliberately not a straight-lining detector — someone who answers “3” to everything is trivially “coherent” and will not stand out on z_rr alone. That consistent form of carelessness is caught instead by the auxiliary indices reported next to it (LongString, IRV). This is why the output is designed to be read as a small panel of complementary indicators, not a single verdict.

Handling real questionnaires

Reverse-keyed items are detected and aligned automatically from the sign of the sample correlation, so you needn't recode anything first. Items on different Likert ranges are rescaled to a common proportion before correlating, so a 1–7 scale can't drown out a 1–4 one. Respondents who gave a single constant answer (zero variance) are handled explicitly rather than producing undefined scores. And a structure diagnostic — calibrated against pure sampling noise — checks that your questionnaire really contains enough strongly-correlated item pairs for the method to have signal, warning you when it doesn't.

When it works best

The signal grows with the number of correlated item pairs available, so ReReRe is strongest on multi-construct batteries. In our validation study detection becomes reliable from roughly 60 items and 60 respondents upward and keeps sharpening as the questionnaire lengthens. Below that it still ranks respondents usefully, but the flags should be read as exploratory — which is exactly what the applicability guide above reports for your specific dataset.

Auxiliary indices in the output — IRV (within-person response variability), LongString (longest run of identical consecutive answers, in file column order), Person-Total correlation and per-row missingness — complement z_rr: careless responding is best judged with multiple indicators.

Privacy: your data never leaves your device — by design

Questionnaire data is often sensitive (personality, mental health, workplace surveys). This tool is built so that uploading your data is not just avoided — it is impossible: there is no server to receive it.

The static files are served by Cloudflare Pages; like any web host, Cloudflare sees the ordinary page request (your IP requesting the site) — but never your data, which is opened only after the page is already on your machine. Your institution's policies for handling data on your own computer still apply.