I have an unusual scale with six items that are sequential in the severity of latent construct. Responses to subsequent items become rarer. Typical CFA works with the first three items. Dropping the next three items could lose some info, as there are sparse responses to these items. Does anyone have any suggestion for dealing with this kind of scale?
A. You yelled or screamed at them. B. You pushed, shoved, grabbed, slapped, or threw something at them. C. You punched, hit, kicked, bit, or slammed them against a wall. D. You beat them up or choked, strangled, burned, or scalded them on purpose. E. You threatened them with a knife or gun. F. You used a knife or gun on them.
Response choices for each item (times): 0 1 2 3-5 6-10 11-20 20+
Sounds like a Guttman scale which factor analysis may have problems with. Does anyone have experience with this?
Jon Elhai posted on Tuesday, October 09, 2007 - 1:06 pm
Some authors in the traumatic stress literature have presented the argument that traumatic experiences (such as the violence items on this "unusual scale") should not be factor analyzed. I believe a relevant paper that discusses the arguments for this position is: Netland, M. (2005). Event-list construction and treatment of exposure data in research on political violence. Journal of Traumatic Stress, 18, 507-517.