In our Rasch analysis of the Symptom Checklist-90-R (SCL90R) using an American outpatient sample (n=298), Elliott et al. (2006) noted among other things that there was a sampling gap in the upper extreme range of the measured underlying client distress dimension. In this paper we remedy this gap by replicating our earlier results with a Scottish inpatient sample (n= 135). When analyzed separately, a three-point rating scale again appeared to provide a better trade-off between reliability and simplicity than the standard 5-point scale. The inpatient sample showed comparable person separation to the outpatient sample (5.13 vs. 5.31), but as expected because of greater participant homogeneity, item separation was substantially lower (4.76 vs. 8.89), indicating less discrimination among items, but equivalent discrimination among persons. Interestingly, when we collapsed the rating scale into only two points ('not at all' vs. all other ratings), there was very little loss of information. Ironically, this suggests that the SCL90R could be used as a 'checklist' after all, in which symptoms are rated either present or absent.
|Publication status||Unpublished - Mar 2009|
|Event||SPR (UK) Ravenscar research Conference 2009 - Ravenscar|
Duration: 14 Mar 2009 → 16 Mar 2009
|Conference||SPR (UK) Ravenscar research Conference 2009|
|Period||14/03/09 → 16/03/09|
- Rasch analysis