Hot answers tagged

8

Your model is saturated. You have 6 covariance moments (three variances and three covariances), and you are estimating 6 parameters (factor variance, two loadings, three measurement error/unique variances). The problem has an exact solution, and zero objective function means that Amos successfully found that solution. It also means that you cannot use the ...


6

In general, structural equation modelling (SEM) with all observed variables is typically called path analysis. One of the main motivations for SEM is to attempt to model relationships between latent variables. By including items rather than the composite score and modelling items as indicators of a latent variable you are able to assess relationships ...


6

I'm paraphrasing your questions a litte bit, in order to still answer them even though I don't think you have the data problems that you think you do: Question 1: Are The Observed Skewness/Kurtosis values acceptable for ML-based SEM? I would say yes. Based on suggested cutoffs for normality that I am familiar with (Skewness > 2, Kurtosis > 7; from Cohen, ...


4

I would strongly recommend that you pick up a copy of Little (2013)--there's simply too much detail to describe about how I would revise your model to accomplish your empirical goals, but Little would be a good (and comprehensive, yet accessible place to start). Here, though, are some broad ideas of how your model development strategy could proceed: ...


4

A CFA with three indicator variables and one latent variable is saturated (it's also called just-identified). It has zero degrees of freedom, and it cannot have a chi-square. It is impossible for the model to be wrong, and the fit is meaningless. It doesn't matter what your data look like, the model will fit perfectly. Yes, they are identified when you ...


4

Main Point With only two observed variables per factor, the latent variable can not generally be estimated. Having correlations between factors presumably adds enough constraints to the model to allow estimation of the latent factors, but without those correlations, the latent factors are not estimatable. What should you do? Add more observed variables ...


4

The covariance is (about) 0.3, which is why that works. So the covariance = 0.3/variance. (The variances are the same, which makes it easier). When you fix the covariance to some other value, it's trying to get the covariance back to 0.3 / variance. It can't change the covariance, because you've fixed that. But by changing the variances, it can get closer....


4

I guess your supervisor is correct and the model is implying mediations. In general, most of SEM and Path analysis involve some mediation or indirect effects. The whole point, of these models is to say, instead of everything being related to everything, i.e. like in a correlation matrix, is to say that some relations are good enough to explain the whole ...


3

I've got a few notes here on the topic of item parcelling. But Jeremy Miles makes the main point. You don't parcel in Amos. You use other software (for Amos users, this is typically SPSS) to combine items into parcels. There are a variety of ways of making parcels, but common approaches simply involve dividing the items in a set into parcels and taking the ...


2

This won't work, the models rarely get inflated in the same way. The proper analysis should involve Satorra-Benter scaled difference. I am not sure as to what the bootstrap analogues would be, although I am sure something can be constructed along the lines of the Bollen-Stine bootstrap (which should have become known as Beran-Srivastava bootstrap). I don't ...


2

If this is truly a scale, and the 11 observed items are endogenous, then your score contains a measurement error, and putting it into a regression model leads to biases: your estimates will be shrunk towards zero (see http://www.citeulike.org/user/ctacmo/article/2663962). This is a poor man's strategy for somebody who has SPSS, but does not have AMOS. If you ...


2

Short Answer Specifying time-scores as 0, 1, 2…, or 0, 0.11, 0.22, 0.33…, or 0, 2, 4, 6… will all let you specify linear models with the same parameter estimates (except the values of any freely estimated time scores) if all time points were measured at equal time intervals. Longer Answer Latent Growth Models (LGM) incorporate time through fixed factor ...


2

Wow. 120 items is a lot. Are they yes/no (0/1)? Do you just use a simple sum? You could try approaching this with fixed reliabilities of your factors. Compute your scores and store them as say f1score through f5score. In AMOS, specify latent variables as circles with f1 through f5 in them, and connect them to the rest of the model. Now, connect each of the ...


2

Yes, the honest thing to do is to present both models. You used SEM in two different ways (described by Joreskog in a chapter in the book "Testing structural equation models", edited by Bollen and Long). You first used a strictly confirmatory approach - this failed, as it almost always does. You then moved to a model generating approach. Ideally, you ...


2

The last question can be peeled off. If I have understood you correctly, you have just cloned the data to get a big enough sample size. That makes no sense at all. Even if you can fool the software, it is still statistical nonsense. Think of it this way: if it is valid, why isn't it prominent in every statistical text as a simple to understand, low ...


2

Just don’t. Post hoc power is meaningless. However, if you insist on it, you can compute it correctly by noting that power is the probability of obtaining statistical significance. Therefore: If H_0 was rejected, post hoc power = 1.0 Otherwise, post hoc power = 0.0


1

Behavioral_intention to actual usage is also good i.e. 0.32 but the real problem is that the R-squared correlation value for actual usage is only 0.10 which is a matter of concern for me. $r$ (correlation) between behavioral intention and actual usage is $\approx 10$. $R^{2}$ (coefficient of determination) for (presumably) linear regression of actual ...


1

IIRC, parcelling is a pragmatic approach to latent variable analysis wherein the construct of a factor score is done by simply taking a sum score. If that's what you understand parcelling to be, it is fully possible to create such a sum score from a matrix of means and covariances. You simply sum the means and multiply by the number of factors to obtain the ...


1

Yes, you can do that. I don't think it's the traditional way. It won't change your analysis or results in any meaningful way. You probably shouldn't do it.


1

If you have a theory, then you should test it with CFA. The problem with EFA is that there are an infinite number of solutions, all of which are (statistically) equally good when (talking about rotation). There is no point testing a second order factor with only two indicators - that model is exactly equivalent to a two correlated factor model.


1

As far as I know, measurement invariance testing is usually performed in SEM context, when research sample contains multiple groups. In SEM context, measurement invariance is often referred to as factorial invariance. It is definitely a good idea to perform both measurement invariance analysis as well as common method bias analysis prior to creating ...


1

Yes, you would test whether each of the parameter estimates between the latent variables are 0. I would first test them as a group, and then individually. You will get p-values for each of them in this model, but I would also estimate models where you constrain each of them to 0, and all of them to 0, that way you can look at all the different fit statistics:...


1

Don't use a composite variable in a structural equations model. Instead, add a latent variable that the manifest variables (i.e., the items) all measure. The composite variable isn't what the items measure, it is just (presumably) a better measure of the latent variable in question. If you were to add a composite variable, it would decrease the ability of ...


1

AMOS can run logistic regressions but I obtained different results for a logistic regression conducted in SPSS AMOS or conducted in SPSS. This might be according to your AMOS version and license. To be sure, I focus on SPSS. For more information - Intro to AMOS Bayesian SEM and MCMC


1

Yes, sort of, but only using the Bayesian approach. Use lavaan, an R package, and you can do this. You can also do this using an interaction term (or set of interaction terms) using the logistic regression function in SPSS.


1

You have two mediation paths in your model: Innovativeness affects Outcome (a) directly and (b) through Ease of Use and Attitude; Usefullness affects the Outcome (a) directly and (b) through Attitude. An effect is said to be fully mediated by the indirect paths if the direct coefficient is zero provided all other relevant paths are in the model. AMOS ...


1

In AMOS there is a plug-in that calculates SRMR for you. Across the top there is a "Plugins" drop down menu, and the SRMR plug-in is there.


1

As the model stands, it is effectively a model with one common factor (Models of Attachment) and three indicators. Such a model is just identified (0 DF) unless there is more to the model. So besides imposing constraints on this model, think about expanding the model to include other variables. You might be headed that way already, in which case the current ...


1

This doesn't directly answer your question, since you are using AMOS, but I wanted to point out that there are two approaches to SEM. One uses an approach based on covariances and maximum likelihood, the other uses an approach based on the original data and partial least squares (PLS). I believe AMOS (LISREL, Mplus, etc) use the first approach, which is ...


1

For many of the most common applications in SEM, analyzing the raw data is equivalent to analyzing a covariance / correlation matrix. In these applications, one of the first things your computer program will do is calculate a covariance / correlation matrix. I should note, though, that analyzing a covariance matrix is not equivalent to analyzing a ...


Only top voted, non community-wiki answers of a minimum length are eligible