Structural Equation Modeling with Small Samples: Test Statistics

Abstract

Structural equation modeling is a well-known technique for studying relationships among multivariate data. In practice, high dimensional nonnormal data with small to medium sample sizes are very common, and large sample theory, on which almost all modeling statistics are based, cannot be invoked for model evaluation with test statistics. The most natural method for nonnormal data, the asymptotically distribution free procedure, is not defined when the sample size is less than the number of nonduplicated elements in the sample covariance. Since normal theory maximum likelihood estimation remains defined for intermediate to small sample size, it may be invoked but with the probable consequence of distorted performance in model evaluation. This article studies the small sample behavior of several test statistics that are based on maximum likelihood estimator, but are designed to perform better with nonnormal data. We aim to identify statistics that work reasonably well for a range of small sample sizes and distribution conditions. Monte Carlo results indicate that Yuan and Bentler's recently proposed F-statistic performs satisfactorily.

Securing Retirement at a Young Age. Exploring the Intention to Buy Longevity Annuities through an Extended Version of the Theory of Planned Behavior

“…Although the most common estimation method within CFA is maximum likelihood (ML), we used an asymptotically distribution-free (ADF) estimator [59], which, by taking into account the kurtosis of the observed variables, avoids the assumption of multivariate normality (The normality assumption is usually not met when the observed data are discrete (as occurs when using ordinal scales). The ADF estimator requires very large samples to behave like a χ 2 [60], and models with more than 20 variables are not feasibly estimated [61]). We adopted this estimation method while aware that it could have had an effect on the incremental fit indexes (CFI and TLI), lowering their value [62].…”

Measuring Meaning in Life

“…Non-normal estimation methods were employed to assess the fit of each model: the scaled chi-square (S-BX2; Satorra and Bentler 1994), the robust incremental fit index, and the robust comparative fit index (IFI and CFI;Bentler and Yuan 1999). A non-significant chi-square and values greater than .90 for the IFI and CFI reflect good model fit, and values between .85 and .90 reflect moderate model fit (Jo¨reskog and So¨rbom 1993).…”

Executive Function Associated to Symptoms of Attention Deficit Hyperactivity Disorder and Paediatric Bipolar Disorder

“…However, the use of SEM in small samples is a valid method according to the literature (Baker, 2007;Bentler & Yuan, 1999). Small samples tend to reject the right models as mentioned by Hu andBentler (1999, cited in Brown, 2006):"TLI and RMSEA tend to falsely reject models when N is small" (p. 86).…”