I am estimating a structural equation model where the dependent variable is the binary variable (measured with Yes or No). Other independent variables are latent constructs measured through mulitple indicators. I am using SMARTPLS, can any one suggest which method is used to estimate this model with binary outcome variable?
Dear Rehman,
PLS is the method adequte to estimate you model. Also you will can use logistic regression.
To use the SmartPLS you can adote the PLS Multi-Group Analysis in SmartPLS 3.0:
Is very easy. See the tutorial:
https://www.youtube.com/watch?v=qssDqR_-Q7Y (there are others in YOUTUBE).
hope this help you
Best regards
Dirceu
Dear Rehman,
PLS is the method adequte to estimate you model. Also you will can use logistic regression.
To use the SmartPLS you can adote the PLS Multi-Group Analysis in SmartPLS 3.0:
Is very easy. See the tutorial:
https://www.youtube.com/watch?v=qssDqR_-Q7Y (there are others in YOUTUBE).
hope this help you
Best regards
Dirceu
Thank you for your reply. I appreciate your prompt response.
Could you further guide, what will be interpretation of the MGA in SmartPLS? I mean my DV is binary (with Yes and No). Should I break down data based on the responses (Yes and No) and carry out the MGA.
Pls employs non parametric approach , so there is no need to break down data.
You don't want to pretend that your dichotomous DV is continuous, so you really need a logistic regression option. Maybe your PLS package offers such an option. If not, then you migth use a logistic regression package and use the observed variabls to directly predict the dichotomous outcome variable.
Together with my co-authors, we developed an approach to PLS that can take the categorical nature of a binary variable into account, see the following:
Article Partial least squares path modeling using ordinal categorica...
It can be easily employed in matrixpls an R package for PLS.
I disagree with Imran, that the scale of your variables can be ignored in PLS. I know that it often happens and that the error introduced is quite small for variables with more than 5 categories. However, for variables with a smaller number of categories, the bias is not neglectable. PLS is based on OLS regression, therefore an implicit assumption is that the variables are metric scaled...
Best regards,
Florian
me too having this problem, but i am using AMOS 23 version, can any one help me ? how to do binary logistic model with AMOS?
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Structural Equation Modeling