I have one continuous independent variable and four continuous dependent variables.There is moderate correlation strength between this four dependent variables(0.6 Multivariate in SPSS, with my IV listed as "covariate" and my DV under dependent varaibles. I only get four corrected beta and it's significance. I don't know how to compare this four beta then...
1. I am not familiar with those method. So whether or not that my initial plan can be accept? 2. If I must do MANOVA, What should I call it and what assumptions to test(I don't even know how to google it...). Note that it described as being a MANOVA, however, there is NO fixed factor/group/category variable used in the analysis. 3. If I do the right thing in SPSS, how can I get the statistic difference between the four beta value...
Any comments will be appreciated!
Dear Fan,
First of all, what is the statistical question you would like to answer?
There is nothing in data that forces us to think a certain way. It is the other way around.
Now, if the interest in your 4 DVs are separate - as is, you plan to look at all of the relationships between IV and each DV on its own because these represent meaningful and important answers to you - then you can use 4 univariate tests for this.
If your interest is in whether their is a joint effect (across all 4 DVs but you are indifferent to which) or you wish to compare the coefficients across DVs then a multivariate test might be useful. SPSS can only handle multivariate regression within the MANOVA command which is why you were getting that advice. MANOVA does have a number of assumptions and requirements. GLLMMs can handle much of what MANOVA cannot.
What about SEM, PLS, and other alternatives? You have a very simple situation so unless you have a highly complicated hypothesis you probably don't need either one. That said, they can tell you about the pattern of relationships between IV and all 4 DVs just as the MANOVA
Consider going to a good site with examples (e.g. http://www.ats.ucla.edu/stat/dae/ )
to see if you can make up your mind a little bt about what you want to know.
Good luck.
Bruce gave good advice. However, I always found SPSS confusing. Here is the same information that Bruce suggested getting for SPSS only for the R package. If this seems more confusing then I would stick with SPSS. The pdf file is the one that tells you about MANOVA. As Bruce said you need to know the question that you want to answer before all else. Best wishes.
http://www.amazon.com/gp/product/0321888030/ref=s9_simh_gw_g14_i1_r?ie=UTF8&fpl=fresh&pf_rd_m=ATVPDKIKX0DER&pf_rd_s=desktop-2&pf_rd_r=04RJXSKQQQ81AJ8CMR68&pf_rd_t=36701&pf_rd_p=2091268722&pf_rd_i=desktop
Dear Bruce,
Sorry for my late reply. I spend some time to understand those different method and I think I find the way to get started due to your clear answer.
Actually, I want to know whether my IV can predict those four slightly different DVs indifferently. I have completed the multivariate regression and did Wald test for post coefficient compare yesterday. The result told me that there is no statistic difference between those coefficients.
Thank you very so much for your help.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS