We compared the composite scores of the indirect and direct measures (attitude/subjective norm/perceived control). They were all strongly correlated. But in some journal articles, authors seem to have compared individual items in the indirect measures to the direct measures. Hence, since correlation between composite scores are good, do we assume that all questions in the indirect measure capture the theme seen in the direct measure? Or is there a way to individually compare each item in the indirect measure to the direct measure?
Since direct and indirect measures of attitudes seem to always be correlated, I wonder why people still use both measures. I am always looking for the fewest number of items I can use to measure a construct to minimize subject burden. I prefer the direct method myself - fewer items and easily understandable for people.
I would suggest you try path analysis. Path analysis is used to estimate a system of equations in which all of the variables are observed. Unlike models that include latent variables. These models assume perfect measurement of the observed variables and only the structural relationships between the observed variables are modeled. This type of model is often used when one or more variables is thought to mediate the relationship between two others (mediation models). Similar models setups can be used to estimate models where the errors (residuals) of two otherwise unrelated dependent variables are allowed to correlated (seemingly unrelated regression), as well as models where the relationship between variables is thought to vary across groups (multiple group models).
Since direct and indirect measures of attitudes seem to always be correlated, I wonder why people still use both measures. I am always looking for the fewest number of items I can use to measure a construct to minimize subject burden. I prefer the direct method myself - fewer items and easily understandable for people.
Thanks so very much for your answers. I prefer to use both measurement to make sure that the questionnaire is valid - to see if the theme in the direct measure is captured in the indirect assessment. But yes I agree - direct measures are simpler and less burdensome for subjects.
Rabin
Hello,
if your sample is big enough you may use Structural Equation Modeling. Here you define latent constructs by a measurement model. The distinction of direct/indirect measures might explain unique variance. This could be operationalized by a so called "Method Factor" within the measurement model.
Some literature on this:
Pohl S, Steyer R. Modeling common traits and method effects in multitrait-multimethod analysis. Multivariate Behav Res 2010;45:45-72.
Greetings
Daniel
Hello Yogarabindranath,
I have a accumulated a lot of TPB studies from a range of journals and applications. Perhaps some comparative data will be of interest to you.
See attached:
Daryn
Which item to remove from a set of items measuring the latent constructs? And you also add concerning the correlation between constructs.
To solve this problem, I suggest you carry out the pooled-CFA combining those constructs. If the correlation between constructs is high, then you identify which item/items in these two constructs are highly correlated. You can identify a pair of redundant item using Modification Indices. High MI (normally above 15) indicates that particular pair of items are redundant. You could delete the lower factor loading between the two.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Applied Statistics
- Survey Design
- Questionnaire