1
$\begingroup$

I am trying to run a SEM to see the causal relationship among latent variables. I ran the CFA and SEM using Amos, but I realized that the model is very complicated since one factor, which I originally thought to be unidimensional, is two-dimensional, and that makes the model look very messy. Since that factor is not derived from a theoretical background, it seems hard to force to use it as unidimensional. Also, a construct validity test also indicated that the factor is two-dimensional.

I know that creating parcels and running the path model has lots of limitations, such as not including error, and it is not necessarily used to create a parsimonious model, but I want to compare whether there are any big differences between CFA and a model with item parceling.

I saw lots of articles that have debated item parceling, but couldn't find how to conduct item parceling in Amos. How can I conduct item parceling in Amos? Thanks!

$\endgroup$
  • $\begingroup$ You create the parcels before you do the SEM, so you create parcels by summing or averaging items in SPSS. $\endgroup$ – Jeremy Miles May 27 '14 at 2:20
  • $\begingroup$ Hi Jeremy. Thank you for your response. So, you mean that go to "Transform"->"Compute variable" and either create summing items variables or mean items variables? Which is more common to use? To me, mean items seem more make sense.Last step seems after creating these new variables, treating them like observed variables...I can see that why item parceling is so controversial..Several articles have mentioned that I should not use item parceling if the EFA result indicated that factor is not unidimensional..can I ask why it should be unidimensional? $\endgroup$ – user46203 May 27 '14 at 3:02
3
$\begingroup$

I've got a few notes here on the topic of item parcelling.

But Jeremy Miles makes the main point. You don't parcel in Amos. You use other software (for Amos users, this is typically SPSS) to combine items into parcels. There are a variety of ways of making parcels, but common approaches simply involve dividing the items in a set into parcels and taking the sum of mean of items within parcels to form the parcels.

In SPSS, you can often do this efficiently by using syntax. E.g., Something like this:

compute parcel1 = mean(i1, i2, i3).
compute parcel2 = mean(i4, i5, i6).
execute.
$\endgroup$
  • $\begingroup$ Hi.Thank you for your comments. I've visited your website which was very helpful. Now I figured out how to conduct item parceling and path analysis using SPSS and Amos. My another question is that whether the factor that was not a unidimensional in EFA result can be used to create parcel. Several articles have mentioned about this issue and I wonder whether the factor must be unidimensional.. $\endgroup$ – user46203 May 27 '14 at 3:17
  • $\begingroup$ My suggestions would be to split your question into two. i.e., this question asks about how to create item parcels; then ask a separate question about "How to create item parcels for a factor that is not unidimensional?" $\endgroup$ – Jeromy Anglim May 27 '14 at 3:20
  • $\begingroup$ I apologize for the confusing of my questions. I am not a native speaker and that might make the questions looks complicated. $\endgroup$ – user46203 May 27 '14 at 3:27
  • $\begingroup$ It's no problem. It's just a suggestion for you to ask a separate question. This site works best where each question answer one discrete question. That way, when people google in the future for answers to a specific question, they'll find something that matches their needs. $\endgroup$ – Jeromy Anglim May 27 '14 at 4:20

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.