Hello,
Using secondary and continuous data, i am trying to run a SEM with two latents. One indicator has both negative and positive values. Is it wrong? (I am gonna run the model in AMOS)
Thx on advance
Hi Ceren,
the meaning of the loading depends on both--the meaning of the latent variable and the metric and its meaning of the indicator.
It is hard to understand what the latent variable is supposed to be when one of its indicators is a difference score. And yes, calculating difference scores has been discussed as a bit problematic.
Edwards, J. R. (2001). Ten difference score myths. Organizational Research Methods, 4(3), 265-287.
Edwards, J. R. (1995). Alternatives to difference scores as dependent variables in the study of congruence in organizational research. Organizational Behavior and Human Decision Processes, 64(3), 307-324.
Can you tell more about both latent variables and ALL indicators (x's and y's)? Perhaps having change in equality as a latent difference score model would be an alternative:
McArdle, J. J. (2009). Latent variable modeling of differences and changes with longitudinal data. Annual Review of Psychology, 60, 577-605. doi:10.1146/annurev.psych.60.110707.163612
I did that the first time in spring 2020 and it is doable albeit not immediately intuitive.
Beyond that, having a model with just two latents makes huge assumptions that there are no unmeasured common causes of both groups which makes only sense in randomized trials....
Best,
--Holger
Hello Ceren,
Yes, you may use variables for which negative scores are legitimate values, as well as variables for which scores may be negative or positive.
Good luck with your work.
Hello Ceren,
I actually don't understand what you mean with "negative values".
As always: A detailed description of the model, background and question wordings would be so helpful, see this and further posting tips:
https://www.researchgate.net/post/Netiquette_proposal_for_questions_on_structural_equation_modeling
Best,
Holger
@Ceren
Not getting your point exactly whether you're talking about the negative factor loadings or path coefficients. Kindly elaborate a little about with a screenshot of your model in AMOS.
Holger Steinmetz Imran Anwar Thnx for your interest. I attached a simple represantation of my model. Indicator y4 is a differencial value. More precisely, it is a value calculated as "income inequality in 2018- income inequality in 2008". As you can guess, it takes negative values for some observations, positive for some other. I subtracted income inequality variables for two years and created a new variable representing change in income inequality. Is it ok? Thnx on advance
Hi Ceren,
the meaning of the loading depends on both--the meaning of the latent variable and the metric and its meaning of the indicator.
It is hard to understand what the latent variable is supposed to be when one of its indicators is a difference score. And yes, calculating difference scores has been discussed as a bit problematic.
Edwards, J. R. (2001). Ten difference score myths. Organizational Research Methods, 4(3), 265-287.
Edwards, J. R. (1995). Alternatives to difference scores as dependent variables in the study of congruence in organizational research. Organizational Behavior and Human Decision Processes, 64(3), 307-324.
Can you tell more about both latent variables and ALL indicators (x's and y's)? Perhaps having change in equality as a latent difference score model would be an alternative:
McArdle, J. J. (2009). Latent variable modeling of differences and changes with longitudinal data. Annual Review of Psychology, 60, 577-605. doi:10.1146/annurev.psych.60.110707.163612
I did that the first time in spring 2020 and it is doable albeit not immediately intuitive.
Beyond that, having a model with just two latents makes huge assumptions that there are no unmeasured common causes of both groups which makes only sense in randomized trials....
Best,
--Holger
Holger Steinmetz Thnx for your recommends. My study is on health systems of countries. Health system determinants is exogenous latent and health system output/ quality is endogenous latent. Health system determinants' indicators are nurse number, GDP , income equality etc., and health system quality is to measured by HIV prevalance, mortality rate etc. which are low for improved systems. So, you think two latents are not enough? Thnx on advance
Hi, you are the expert but having GDP and income equality as (causally dependent) reflective indicators of a latente health systems variable is rather strange to me....I would rather guess that economic factors are a cause of the health systems is it depends on ecomic resources....?
Likewise, I find the strategy to view all concrete health systems outcomes as mere reflections of a latent factor strange...Why not model these as singular depentent variables (and again this would allow to model a change score model for the inequality variables.
What is the N and the chi square test of this model?
With regard to the "enough" question: This all depends on whether there are unobserved confounders of a certain relationship that have to be controlled.
Best,
Holger
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Structural Equation Modeling