AtheneaLab

Data Cleaning

Upload your survey dataset in Excel and get the clean dataset, the coded dataset and a log of the process, with the text to report it in your article.

Your data never leaves your computer. The whole process runs inside your browser: AtheneaLab does not receive, upload to any server or store your dataset.

1

Upload your dataset

Excel (.xlsx) or CSV from Microsoft Forms, Google Forms or any other source.

2

Initial review

Columns are identified and the number of cases each filter would affect is calculated.

3

Your decisions

You decide which procedures to apply and choose the age range.

4

Download

Clean dataset, coded dataset and a PDF log.

Methodological decisions

Identified columns

    Variables and items

    Each question is a card: drag it to its variable (or use “Move to…”) and, if you wish, split the variable into dimensions. In the coded dataset each item is named with its variable prefix, dimension and question number (V1_D1_1, or V1_1 without dimensions). Mark reverse-scored items to recode them; the ones the tool suggests arrive checked and highlighted. “Check” sets which variables are used to check incomplete responses and invariant responding.

    Counts are sequential: each step is applied to the records that remain after the previous one.

    Which procedures are applied and why?

    An online survey always includes responses that do not reflect what the person thinks: someone who answered without reading, left the questionnaire halfway or did not agree to participate. Data cleaning removes those cases with explicit, documented criteria, so that results are reliable and other researchers can replicate the process (Curran, 2016). The procedures are applied in this order:

    1

    Informed consent

    Why: only data from people who agreed to participate can be analyzed. It is an ethical requirement of all research with people.

    Anyone who did not answer "Yes" to the consent question is excluded.
    2

    Incomplete responses

    Why: without all the items the scale score cannot be calculated, and incomplete cases bias the analyses.

    Records with at least one unanswered scale item are excluded.
    3

    Responses that were too fast

    Why: someone who answers in very little time could not read the questions; speeding usually comes with repeated patterns (Zhang & Conrad, 2014).

    Duration below one third of the median. Only if the dataset has start and completion times.
    4

    Responses that were too slow

    Why: an extreme time suggests interruptions or that the person left the questionnaire and came back later, without continuous attention.

    Duration above Q3 + 1.5 × IQR (Tukey's outlier criterion). Decided separately from fast responses.
    5

    Invariant responding (straight-lining)

    Why: choosing the same option on almost every item indicates that the statements were not read (Krosnick, 1991; Meade & Craig, 2012).

    ≥ 80% identical answers in at least one variable, assessed with all its items. If a variable has few items or the filter flags too many people, you are warned before applying it.
    6

    Age

    Why: impossible ages or ages outside the study population are typing errors or people who do not belong to the sample.

    You choose the range before running.
    7

    Attention check item (optional)

    Why: a question with an instructed answer ("select Agree") detects people who answer without reading (Meade & Craig, 2012).

    Only if your questionnaire includes one; you indicate the correct answer.
    8

    Coding

    Why: statistical software works with numbers; each answer gets a documented code so the analysis is transparent.

    Agreement and frequency Likert 1–5, Yes/No 0–1, sex, marital status, education level and other categories with standard codes; multiple-choice questions become 0/1 columns. Items you mark as reversed are recoded (on a 1-to-5 scale: 1 ↔ 5, 2 ↔ 4) and each item gets a short code such as V1_D1_1. The tool suggests reversed items: items with an item-rest correlation below −.20 within their variable arrive checked and highlighted for you to review.

    What do you get?

    Clean dataset

    The same file with all its original columns, but only with the valid records. Useful for review and archiving.

    Coded dataset

    Ready for the software of your choice, such as jamovi or SPSS: identifier, duration in minutes, items with short codes (V1_D1_1) and variables as numbers, plus a Dictionary sheet with the wording of each code. Emails, names, timestamps and open-ended text are removed.

    PDF log

    Each procedure with its rationale and decision, the figures for each step, the flow figure, the code table and an APA-style guide paragraph for your Method section, which you should paraphrase.

    References

    Curran, P. G. (2016). Methods for the detection of carelessly invalid responses in survey data. Journal of Experimental Social Psychology, 66, 4–19. https://doi.org/10.1016/j.jesp.2015.07.006

    Johnson, J. A. (2005). Ascertaining the validity of individual protocols from web-based personality inventories. Journal of Research in Personality, 39(1), 103–129. https://doi.org/10.1016/j.jrp.2004.09.009

    Krosnick, J. A. (1991). Response strategies for coping with the cognitive demands of attitude measures in surveys. Applied Cognitive Psychology, 5(3), 213–236. https://doi.org/10.1002/acp.2350050305

    Meade, A. W., & Craig, S. B. (2012). Identifying careless responses in survey data. Psychological Methods, 17(3), 437–455. https://doi.org/10.1037/a0028085

    Zhang, C., & Conrad, F. G. (2014). Speeding in web surveys: The tendency to answer very fast and its association with straightlining. Survey Research Methods, 8(2), 127–135. https://doi.org/10.18148/srm/2014.v8i2.5453

    How to cite this tool

    AtheneaLab. (2026). Data Cleaning [Web application]. https://athenealab.com/en/data-cleaning/