1

I've read many of the posts on this board about the difference between reliability and validity, and most answers indicate that there is little use for a measure that is not reliable. However, I'm wondering if that has to be the case? That is, could there be some good reason that a measure isn't reliable (e.g., doesn't have good test-retest reliability) but is still useful? Or is it really true that a measure found to be unreliable should be done away with?

user2917781
  • 475
  • 1
  • 5
  • 13

1 Answers1

1

Reliability tells us about internal consistency, i.e. it shows that we are able to replicate the results. Imagine that you measured your child's temperature, after repeating the measurements few times during fifteen minutes you got the following readings: 35.1°C, 36.6°C, 38.1°C, 35.7°C, 37.3°C, 39.9°C - what do they tell you? I'd say that they show that your thermometer is unreliable and cannot be trusted.

Could there be some good reason that a measure isn't reliable? Yes, for example, when you do not have anything better on hand. Another example, you may want to make your test shorter so your study participants do not have to answers multiple-page long questionnaire so not to discourage them, what can potentially make it less reliable than the full test etc. Saying it shortly, sometimes practical reasons make you use less reliable measurements.

Community
  • 1
Tim
  • 108,699
  • 20
  • 212
  • 390
  • Thanks for the comment. After thinking about it more, I was thinking that maybe a less reliable measure could be helpful for still indexing group differences. For example, say that on average people in one culture have higher scores than scores in another culture. Even if the measure is unreliable (any one person's score isn't likely to be the same if they took the measure again), the difference between cultures may still be reliable. Is this intuition correct? – user2917781 Sep 17 '16 at 17:21