Scabies Pimples After Treatment

"Every time we have, we rely on a principle of identity, ie, at least for us the objects we have are identical enough to be interchangeable."

Wright and Mok, 2004


When I read the first chapter of Introduction to Rasch Measurement edited by Everett Smith and Richard Smith (ISBN 0-9755351-1-0) I liked this phrase, and I took the liberty of translating into Castilian. The idea brings me to one of the important problems of the validity of psychometric tests: the one-dimensional. Although this problem is often neglected in the texts of Classical Test Theory is much in the literature on Item Response Theory.

Question: Where do you usually live score future tests?
Answer: the sum of individual scores of the items.

Solo por hablar del caso de los ítemes dicotómicos, la suma, o puntaje directo es el conteo de respuestas correctas o positivas (que van en la dirección del constructo). Y si queremos contarlos, esto implica que “son lo suficientemente idénticos como para ser intercambiable”. Con esta afirmación, los autores están reafirmando la importancia de la unidimensionalidad: si los ítemes son combinados linealmente, es porque comparten algo en común, es decir, miden fundamentalmente lo mismo. ¿Cómo podemos fundamentar la suma de un par de ítemes si estos se refieren a algo muy diferente?

Si los ítemes tienen un grado significativo de varianza común compartida, say that this is due to their covariance is produced primarily by the same latent trait. Even if we make a statement in causal terms, we say that the observed variance in these items is largely caused by the same latent trait.

Of course, there is no perfect one-dimensionality. What we are measuring instruments whose items are not far from this assumption. Some would say, are essentially one-dimensional tests. It is precisely this aspect, which we aim, to take and measure instruments whose scores (and interpretations) are more valid. This is the objective of Rasch models (for dichotomous items, partial credit rating scale, etc..) scale to achieve a set of items that do not stray far from this assumption undimensionalidad, in addition to a number of assumptions, which will write at another time.

Best Regards, Andrew