Reading sources critically · Lire les sources de manière critique
| English | Français |
|---|---|
| operationalisation/ˌɒpəˌreɪʃənəlaɪˈzeɪʃn/ | opérationnalisation |
| authority/əˈθɒrɪti/ | autorité |
| currency/ˈkʌrənsi/ | monnaie |
| bias/ˈbaɪəs/ | un biais |
| missing data/ˈmɪsɪŋ ˈdeɪtə/ | données manquantes |
Two percentages may measure different things
- In a fictional cohort of 100 young people, one indicator counts 8 actively seeking work, while another counts 21 outside work, education or training. The second category is broader.
- These are different indicators, not two interchangeable unemployment rates. Operationalisation 操作化 turns a concept into a defined measure.
In the fictional cohort, 8% are jobseekers and 21% are outside work, education or training. Why are these not interchangeable indicators?
The example states different categories. A real unemployment rate also needs its own labour-force denominator.
Read the source and its scope
- Authority 权威性 concerns expertise and responsibility; currency 时效性 concerns date and relevance; bias 偏差 concerns systematic distortion. None is settled by a logo alone.
- Check the producer, definition, denominator, period and method. A funder’s interest is a reason to inspect a claim, not proof that every result is false.
Match each test to the question it asks. · Reliez chaque test à la question qu'il pose.
Inspect expertise, date, definition, selection and interests for this claim. A source label does not prove correctness or bias.
Investigate missing observations
- Missing data 缺失数据 may have different mechanisms. Nonresponse can relate to the topic, but this must be investigated rather than assumed in every data set.
- An online-only survey may exclude people without access. Record that risk and what is known; do not invent their answers.
Missing data always occur at random.
False. Missingness can relate to the topic or other recorded factors, but its mechanism should be investigated rather than assumed.
Write one question you would ask about a survey reporting that "70% of students feel stressed". · Écrivez une question que vous poseriez à propos d'un sondage rapportant que "70 % des élèves se sentent stressés".
Example: "How was 'stressed' defined in the question, and what were students actually asked?" · Exemple : "Comment 'stressed' a-t-il été défini dans la question, et qu'est-ce que les élèves ont réellement dû faire ?"
Compare like measures
- Two sources using different definitions or periods can give different values without either being false. Equal labels do not guarantee equal measures.
- The fictional 8% and 21% share a denominator but count different categories. A real unemployment rate normally needs its own defined labour-force denominator.
One indicator counts jobseekers and another counts people outside work, education or training. What should a comparison do?
The groups differ. Keep the category and denominator visible before interpreting a difference.
State the next check
- Before comparing two regions, inspect population, age limits, categories, recording practices and dates. A precise decimal is still a measurement of a definition.
- Report the indicator by its actual name and identify unresolved comparability. Definitions are one possible source of disagreement, not the explanation for every disagreement.
Report the indicator by its actual name and identify unresolved comparability. Definitions are one possible source of disagreement, not the explanation for every disagreement.
Before comparing two countries' crime figures, what must you check? · Avant de comparer les chiffres de criminalité de deux pays, que devez-vous vérifier ?
Definitions and recording practices differ, and a comparison built on that is worse than none. · Les définitions et pratiques d'enregistrement diffèrent, et une comparaison basée dessus vaut moins que rien.