Fichas
lecture notes
Fichas del conjunto (9)
- 1measurement error, meaningthe distortion of the true score because the measurement instrument is not perfect
- 24 steps to follow for construct validity1. Define a construct in terms of a (nomological) network of relationships (e.g. intelligence: working memory = positive relationship, learning ability, etc.)
2. Operationalize every construct with a measurement
3. See if the relationships between measurements are aligned with the theory
4. If yes: there is support for the theory in question. If no, then either the test or the theory should be revised. - 3interpretationist account of validityvalidity is an integrated evaluation about the extent to which the interpretation of the test scores is supported by empirical evidence and theoretical arguments
- 4causal account of validitytest X is valid for attribute Y when differences in Y cause differences in X. we have to establish a causal relationship between the construct and the test, only then we can conclude that a test is valid. however, there are many variables you can’t manipulate.
- 55 aspects of items to determine whether they are of good quality- Item difficulty/item attractiveness
- Item variance (omit those with small variance)
- Inter-item correlation
- Item discrimination
- Item reliability index - 6what's a good inter-item correlation?around 0.7. you don't want it to be too high (then they don't add anything) or too low (then you're not measuring the same thing)
- 7explanation of item discriminationHow well can a given item distinguish between people that differ on the underlying
construct? For example, if Pete is slightly more neurotic than Rose, then Pete should score
(e.g.) 4 and Rose 3.
--> how well can an item predict the test score? - 8in a multitrait-multimethod matrix (MTMMM), which type of correlation gives evidence for convergent validity?monotrait-heteromethod correlations
- 9in a multitrait-multimethod matrix (MTMMM), which type of correlation gives evidence for discriminant validity?heterotrait-monomethod correlations (these correlations should be LOW! if they're high, then that evidence against discriminant validity bc the variables correlate when they shouldn't)
Información
- Descripción
- university PTPR Fichas sobre lecture notes, creado por Sabine - el 22/09/2021.
- Creado por
- Sabine -
- Visitas
- 0
- Valoración