Hello,
I have long-term (4 year) observations of hourly soil respiration, temperature, and moisture data. I am interested in linking these observed variables to latent variables of “topography” (using indicators of soil depth, elevation, and curvature or topographic wetness index) and “soil properties” (using indicators of soil carbon and texture) in a structural equation model.
However, while a variable like soil respiration changed every hour, a variable like elevation remained constant and so is only measured once. Is it possible to reconcile these differences in sampling frequencies using a structural equation model? Is there another method more appropriate for this situation?
Thank you!
Marissa Kopp - If you look at aggregating different samples into one for SEM analysis, this article might interest you as it offers a sampling approach concerning that.
Article Relating Patient Satisfaction to Nurses’ Job Satisfaction, J...
Happy to offer more details if that's the case for your research.
Many thanks,
Ali
Dear Marissa Kopp
Generally speaking structural equation modelling (SEM) can have both observed or latent variables to measure an outcome. However, SEM is mostly applied to test a theory of interest to researchers, which applies that they have a picture of a set of latent variables and how they interact with each other. Therefore, it is often referred to as a theory confirmation approach.
So often SEM is applied to confirm behaviour or theory, in other words, the theme of the study would be related to the individual's beliefs or intention to accept/buy/perform a certain action.
Back to your situation, in which as I understood, that you want to link the observed variables of (hourly soil respiration, temperature, and moisture data) with other 2 latent variables, which are topography measured by (soil depth, elevation, and curvature or topographic wetness index) and soil properties measured by(soil carbon and texture).
I am not sure what you exactly what to measure or test, however, the theme of study does not indicate that there is a theory confirmation part being involved.
I guess it tends more toward predicting - e.g. predicting the topography using observed variables and soil properties. Therefore, I think that SEM is not the appropriate method to proceed further.
I am not sure about the appropriate analytical method that should be applied but I am optimistic that other folks are going to have more insights regarding that.
Best Regards,
Belal Edries
Hello Marissa,
may I ask what the time frame of the measured topography variables are? I guess they are time-invariant predictors...That is, I would assume that your variables of the soil characteristics form different time series with varying intervals nested in groups with varying topographical characteristics?
If yes, my hunch would rather be some kind of GAMM (generalized additive mixed model) in which you can model nonlinear relationships among your time series variables and between-level effects of your topography variables on the level and non-linear trajectory of the time series.
A second approach could be to model individual VAR (vector autoregressive) models for each group and add the lagged effects to a data set (which can be easily done using the fable package in R). In a second step, you could use these coefficients (level and/or slopes) as variables in an SEM, potentially modelled as latent variable. I don't know whether this makes sense or has some drawbacks as I myself am currently digging into time series.
Anyway, you would have to create a data set with equal time intervals by up-aggregating the series with a finer graded measurement. Perhaps someone joins the discussion who (really) knows about this stuff :)
All the best,
Holger
Marissa Kopp - If you look at aggregating different samples into one for SEM analysis, this article might interest you as it offers a sampling approach concerning that.
Article Relating Patient Satisfaction to Nurses’ Job Satisfaction, J...
Happy to offer more details if that's the case for your research.
Many thanks,
Ali
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Structural Equation Modeling