Reliability is an indicator against probabilistic failure of a component or system of components. It is only confidence level, but not the absolute. However, the theoretical values of reliability more than 0.9 is considered good. Partly I will agree with redundancy also
In order to be able to make decisions for individual cases you need to have reliable measures. In measures with a low reliability you are adding apples with pears. Computing a sumscore or scale score by adding phenomena that are not even closely related is not advisable.
In measures with a reliability far above 0.90 you are adding the same information over and over again, which is redundant and useless. it is like asking three times the same question: this persons does not breath, this person does not have a heartbeat, this person does not show electrical activity in the brain. this can be replaced by the statement 'this person is dead'.
When a large number of different variables are combined, you will find a high reliabilty score anyhow. even adding 30 variables of random data make a good scale together. in order to compensate for the effect of a numer of variables that is too high you hacve to heck with the Spearman Brown correction. Every scale with more than 10 items have to be checked with this formula. It is not directly found in spss, but you can program the syntax for a Spearman Brown correction is spss yourself.
I would recommend to think theoretically, rather than focusing on thresholds. Ensure that you have a conceptual sound explanation for your constructs.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Probability
- Probability Theory
- Reliability