Message/Author 

yan wang posted on Wednesday, August 19, 2015  6:47 pm



hello prefessor, I want to use the secondorder cfa to analysis the quality of life for old people. From the users guide i know mplus can handle categorical indictors, both binary and ordered categorical(ordinal) But i wonder if ordered categorical(ordinal) indictors such as likert scale(for example ,in my construct, i have indictors such as "do you satisfied with your life", score from 1 to 5£© and binary indictors £¨such as "do you drink",1 means not drink,0 means drink)can be used in a model at the same time. secondly, i am not sure whether or not mplus can only handle ordinal binary indictors. Can nominal binary indictors treated as the same way in mplus? Thanks very much for your assistance. 


You can have a combination of binary and ordinal factor indicators. Binary indicators are treated the same whether they are ordinal or nominal. 

yan wang posted on Thursday, August 20, 2015  4:36 pm



Thanks for your help Linda.But I still have some qustions. (1)I am confused whether one latent variable can have both binary and ordinal factor indicators or not.if the anwser is yes,how does the correlation between them. (2)besides,i am not very clear about nominal and ordinal binary indictors,can you help me distinguish them.for example,i know sex is a nominal indictor,but what about indictors such as" do you drink,1=not drink,0=drink" "do you have a house""1=have,0=dont have". (3)i want to know,like indictors i just mentioned above£¬in fact£¬i want to treat people drink as the reference group for those not drink.Should i creat a dummy indictors or it is already a dummy inditor£¿if the indictors are dummy£¬can those be treated the same as other binary varibles?how can they be explained according to the estimate outcomes.And if a dummy variable has two or more reference groups,how should it be treated. i am looking forward to your reply.thank you. 

yan wang posted on Thursday, August 20, 2015  4:52 pm



And can you recommend me some reference for my questions. thanks a lot! 


1. Yes, one factor can have a combination of binary and ordinal factor indicators. The correlation between them is a polychoric correlation. 2. With two categories, it can be nominal or ordinal but the methodology is the same. 3. You should use the categories as measured. The number of the category is not meaningful as a value. 


See the Topic 2 course handout. There are many references in that. The Agresti book is very good. 

yan wang posted on Thursday, August 20, 2015  6:18 pm



thanks for your help Linda.but i am not understand very vell of your anwser to my third question.can you give me a further expanation£¿ 


You may want to ask that general analysis question on SEMNET. 

yan wang posted on Thursday, August 20, 2015  6:32 pm



when i choose the indictors ¡°do you drink¡±as a binary indictors to measure the lifestyle of a person,if the outcome is significant can i say the person who is not drink has a better lifestyle than those drink£¿ 

yan wang posted on Thursday, August 20, 2015  7:31 pm



I post the question to the semnet,but there has no one responded me yet,so can you help me?I really appreciate for your help! 

Back to top 