Hi Everyone!
I have a sample of 600 participants, that answered 8 items. As my study is a replication (in the original study they used LCA to identify different groups), I am trying to follow it as best as possible. Because I cannot do LCA in SPSS (though if someone can explain to me how to do it in STATA15 it will be great!), the second best option for me to do, according to my supervisor, is run PCA, and use it to identify the different sub-groups in my sample.
I know how to run PCA and interpreting the results, but I do not know how I can allocate from the sample of the 600 participants, based on their scores in the 8 items, to create the two subgroups (because they are loading on two factors) from those results.
Hope someone can help me :)
Hello,
If there are two factors, but you need to divide only into 2 subgroups, it is logical to use only the first factor (with the highest value Eigenvalue and %Total variance). You should get a table with 600 values of "Factor coordinates of cases" in your results. "Factor coordinates of cases" is accurately calculated in STATISTICA, but in SPSS and STATA it may not be calculated, I don't know.
Take only factor 1. Positive values ( + ) "Factor coordinates of cases" will be the first subgroup, negative values ( - ) this will be the second subgroup. Of course, it is better to take into account two factors, but then four subgroups will be obtained on the basis of "Factor coordinates of cases": Factor 1(+values)Factor2 (+), Factor1(+)Factor2 (-), Factor1(-)Factor2 ( -), Factor1(-)Factor2(+).
Also, for your data, you can probably use Сluster analysis, in particular the method k-means, in your task definition 2 subgroups will be 2-means.
You would not want to run PCA separately on both datasets, since the resulting PCA dimensions would not be comparable.
Since actually you want to know how to describe both datasets with least possible dimensions, you will combine both datasets and perform a single PCA on the combined dataset.
Yet a single PCA dimension will be a linear combination of all input dimensions therefore you can not effectively exclude dimensions, but rather "summarize" your data.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS