rpsych

Find RMSEA of a PCA or factor analysis in R pkg:psych

I can no longer find the RMSEA of a factor or PCA analysis in R. See this code:

library(psych)
data("Thurstone")
Thurstone
mod1<-pca(Thurstone, nfactors=2)
mod1$RMSEA
factor.stats(Thurstone, mod1)

Where is this information stored? I used to be able to run mod1$RMSEA and get it quite cleanly.

Solution

  • After reading the help page for factor.stats I suggest

    factor.stats(Thurstone, mod1)$RSMEA

    So what you may have done before was something like

    Fstats <-  factor.stats(Thurstone, mod1)
Fstats$RMSEA

    There’s another function, principal in pkg::psych that sometimes gets different results but I don’t know if the difference includes the RMSEA values.

    Once I was able to do some hacking on something other than my iPhone I found that including the n.obs argument seems to solve the problem:

    factor.stats(Thurstone, mod1, n.obs=500)$RMSEA
     RMSEA      lower      upper confidence 
 0.1789257  0.1622499  0.1965331  0.9000000

    Added note: You can see that the help page for ?pca doesn't actually document the pca function; rather it describes the principal function. And if you look at the code for pca is shows that the "n.obs" needed by factor.stats is set to NA.

    > pca 
function (r, nfactors = 1, residuals = FALSE, rotate = "varimax", 
    n.obs = NA, covar = FALSE, scores = TRUE, missing = FALSE, 
    impute = "median", oblique.scores = TRUE, method = "regression", 
    use = "pairwise", cor = "cor", correct = 0.5, weight = NULL, 
    ...) 
{
    principal(r = r, nfactors = nfactors, residuals = residuals, 
        rotate = rotate, n.obs = n.obs, covar = covar, scores = scores, 
        missing = missing, impute = impute, oblique.scores = oblique.scores, 
        method = method, use = use, cor = cor, correct = 0.5, 
        weight = NULL, ...)
}
<bytecode: 0x5654e4aea470>
<environment: namespace:psych>

    So if principal had been called everything would have come as expected:

    > mod1<-principal(Thurstone, nfactors=2)
> mod1$RMSEA
NULL
> factor.stats(Thurstone, mod1, n.obs=500)$RMSEA
     RMSEA      lower      upper confidence 
 0.1789257  0.1622499  0.1965331  0.9000000