Can i directly use individual 5point Likert items, as independent variables in multiple regression, as these items are formative construct, not reflective-so i can't transform them to a single mean score. I have read that Likert items on 5 or more point scale can be treated as continuous variable..
Further, my dependent variable has 4 items which are again measured on 5 point liker scale, but are reflective in nature, so i will transform them to a single mean value (to be taken as dependent variable) in multiple regression. Is that OK?
Can you do it? Sure. Should you do it? That depends. The key thing here is whether or not your measure perfectly measures what you are after, or at least close enough that the imprecision doesn't matter. For example, most people accept weight (in kg/lbs/stone) as a perfectly measured variable because only minor inconsistencies (such as the scale, the surface a scale is on, etc.) affect this in usually very minor ways.
Likert scale items/tests typically don't fall under this umbrella. We are utilizing a very loose net to capture our construct (e.g. anxiety, anger, etc.). With this inevitably comes imprecision, and this is often why Likert scale items are paired together into a single composite, as this allows multiple items to provide convergence in measuring a single latent construct. This doesn't say that single-item Likert scale items are invalid, but more that one faces more risks in doing so.
There are some single-item tests out there that do in fact provide reliable measures of certain outcomes across studies (see linked articles). But these are usually not as common and require a fair deal of validation across studies to achieve any sense of accuracy.
References:
Lucas, R. E., & Brent Donnellan, M. (2012). Estimating the reliability of single-item life satisfaction measures: Results from four national panel studies. Social Indicators Research, 105(3), 323–331. https://doi.org/10.1007/s11205-011-9783-z
Wanous, J. P., Reichers, A. E., & Hudy, M. J. (1997). Overall job satisfaction: How good are single-item measures? Journal of Applied Psychology, 82(2), 247–252.
- Badges
- Science topic