Issues with saturated model fit indexes

[Ionut Mone]

Ok so i have been having a problem with r not presenting fit indices for my model.

 From what i`ve figured r doesn`t return fit indices because my model is fully saturated as the number of observations are equal to the number of parameters i want to estimate. If i eliminate parameters or i constraint a variable`s variance to be constant (TraditionAndConformity ~~1*TraditionAndConformity) r does return fit indices. 

Now my question is: can i raport my results without the fit indices while mentioning that my model is fully saturated? If yes, do you know of a reference that might be of use?

And if not what types of constraints would be legitimate so that my model wouldn`t be fully saturated anymore?

The code is below

ibrary(semTools)

library(lavaan)

library(Amelia)

library(mice)

r2 <- r[,c("TraditionAndConformity", "Relatedness", "AgencyAndSelfDirectedness", "SelfDirectionActionC", "ConformityRulesC", "AutoRelEthno", "Heteroetno", "SeparatenessEtno", "IOSTOTAL", "dispototal")]

simpleMediation <- '

Heteroetno ~ b * TraditionAndConformity + c * dispototal

TraditionAndConformity ~ a * dispototal

indirect := a * b

total    := c + (a * b)

TraditionAndConformity ~~1*TraditionAndConformity

Heteroetno ~~ Heteroetno

dispototal ~~ dispototal

'

require(lavaan)

fit <- sem.mi(model = simpleMediation, data = r2, m = 40, miPackage = "Amelia", seed = 12345)

show(fit)

summary(fit, se = TRUE, ci = TRUE, level = 0.95,

        standardized = TRUE, rsquare = FALSE, fmi = FALSE, header = TRUE,

        scale.W = TRUE, asymptotic = FALSE, add.attributes = TRUE)

coef(fit, type = "free", labels = TRUE)

vcov(fit, type = c("pooled", "between", "within",

                      "ariv"), scale.W = FALSE)

anova(fit, h1 = NULL, test = c("D2"),

      pool.robust = F, scale.W = FALSE, asymptotic = F,

      constraints = NULL, indices = TRUE, baseline.model = NULL,

      method = "default", A.method = "delta", scaled.shifted = TRUE,

      H1 = TRUE, type = "Chisq")

[Mauricio Garnier-Villarreal]

Yes, you can report the results of your model, and explain that you dont have fit indices because you have a saturated model. If a reviewr argues you can say that the actual fit indices would be perfect CFI=1, RMSEA=0. You can reference Kline book about what  saturated model is. I had to add a couple of lines to explain in a paper due to a reviewer

In path analysis models, I actually dont like to talk about fit indices. I see them more relevant to the measurement model with latent constructs, when you try to defend that a specific factorial structure is a plausible data generating model for your data. With path, I dont wee the representation of a plausible generative model as useful. Also, a path analysis its easy to make it a saturated model and have "perfect fit".

[Ionut Mone]

Thank you very much for the answer!

--
You received this message because you are subscribed to the Google Groups "lavaan" group.
To unsubscribe from this group and stop receiving emails from it, send an email to lavaan+unsubscribe@googlegroups.com.
To post to this group, send email to lav...@googlegroups.com.
Visit this group at https://groups.google.com/group/lavaan.
For more options, visit https://groups.google.com/d/optout.

--

Mone Ionut-Sergiu

PhD Student

Developmental Psychology Laboratory

Department of Psychology 

Babes-Bolyai University 

400015 Republicii 37 

Cluj-Napoca, Romania 

web: www.psychology.ro