A topic in research methodology
Latent variables
Some things that we might wish to measure in education can be readily observed. It is possible to find the number of pupils present in a classroom at any particular time by counting them. If an observer wanted to find out how much of a lesson was spent in the teacher working with the class as a whole, and how much time was spent on other types of activity (individual work, paired and group work) then this can be found by observing and recording. This requires the observer to interpret situation, and there will be a limit to the precision of timings, and some skill in using a stop-watch would be needed, but basically only one measure is needed for the duration of each activity.
If two different observers watched the same class making these timings, and their observations disagreed (more than an accepted measurement error) then there might be a question of a lack of inter-rater reliability, but this would not be because more than one set of measurements was needed in principle.
Yet researchers often also wish to measure things that cannot be directly observed: attitude to school, self-confidence in mathematics learning, trust in the examination system…
Constructs that cannot be directly observed, such as attitude to school, are which have to be measured indirectly are sometimes called latent variables. (Latent mean hidden – as in the 'latent heat' of freezing/solidification which my lead to a perceived warming when snow melts or the latent heat of boiling – which is why being scolded by steam can do so much damage as a lot of energy is released as the steam condenses.)
Scales are often constructed to help indirectly measure latent variables.
Scale items
For example, someone may write (and then pilot and modify) a set of statements to be rated by learners to measure their attitude to school. A scale is used as it is beleive no one item will directly access the variable of interest, but an array of items may collectively allow an approximate measurement of the variable to be inferred. Perhaps school children are asked to respond to a set of eight statements considered to link to their attitude to school, and the pattern of responses used to infer their relative attitudes to school.
Internal consistency
In a good scale each item contributes to finding out about the latent variable, but each item elicits a somewhat unique pattern of responses. When the scale is used with a sample form the population of interest, it is possible to measure the internal consistency of the scale – the extent to which the different items seem to be 'getting at' the same underlying thing (which if the items have validity, is hoped to be the latent variable).
(Read about 'Internal consistency')
My introduction to educational research:
Taber, K. S. (2013). Classroom-based Research and Evidence-based Practice: An introduction (2nd ed.). London: Sage.
