I want to predict the changes in multiple variables with change in one independent variable in my analysis. Also suggest how to compute using SPSS.
Are the multiple variables at different times (and then may have for example autocorrelated errors), or related in some way. Multivariate ANOVA or time series (longitudinal methods) may be appropriate but more information is needed.
Are the multiple variables at different times (and then may have for example autocorrelated errors), or related in some way. Multivariate ANOVA or time series (longitudinal methods) may be appropriate but more information is needed.
All the dependent variables are sub-scales of a construct, summing up their scores finally provide the total score of a construct.
It looks like a factor, so check for its reliability first, e.g. by using Cronbach's Alpha. Otherwise summing scales would be meaningless.
You can go for multiple regression analysis and develop the model including only those outcome variables which statistically significantly predict your predictor independent variable.
SPSS has option to perform the same.
Thanks and Regards,
Shuchi
By definition multiple regression model predicts the value of a dependent variable (i.e. one outcome variable) based on the value of two or more independent variables (i.e multiple predictor variables). However if I have understood you data correctly you seem to want to do the opposite i.e. predict many outcome variables based on one predictor variable. In such case why not try univariate analysis like Daniel pointed out and test each outcome separately. Unless you wanted to adjust for all the outcomes in one model, then that will be a different case.
Regards
Have a look at
https://www.researchgate.net/post/Can_you_share_any_examples_of_the_use_of_multivariate_regression
Note this is multivariate multiple regression and not multiple regression, that is multiple responses.
I think that Huberty & Morris (1989) give sound advice. They say that if you have a multivariate question, use a multivariate model; but if you have a series of univariate questions (i.e., questions about each outcome variable separately), then estimate a series of univariate models. HTH.
https://pdfs.semanticscholar.org/ee6c/77c99c8e4530d0cccaedf85ed525fb22a02d.pdf
Bruce
I agree with two related caveats - first you may be able to increase statistical power with a multivariate multilevel that borrows strength from one response to another second this can be helpful when response values are missing (conditionally at random)
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS