Model did not converge

[Ainur Amira]

Hi everyone,

I'm the beginner in R and lavaan, Can anyone help me give some suggestion?

I try to run using a small sample size, (let say n=30), but the model only converges at n=55, but when I try using n=76, the model did not converge again. What should I do?

beta <- matrix(NA, nrow = 500, ncol = 4)

FIT <- matrix(NA, nrow = 500, ncol = 7)

var <- matrix(NA, nrow = 500, ncol = 15)

set.seed(12345)

for (i in 1:500){

data2 = data.f(76)

model <- '

KSI1 =~ X1 + X2 + X3

KSI2 =~ X4 + X5 + X6

KSI3 =~ X7 + X8 + X9

KSI4 =~ X10 + X11 + X12

ETA1 =~ Y1 + Y2 + Y3

# REGRESSION

ETA1 ~ KSI1 + KSI2 + KSI3 + KSI4 '

fit <- sem(model=model, data = data2)

beta[i, 1:4] = parameterEstimates (fit, ci = FALSE, standardized = TRUE)[16:19, 4]

FIT[i,1:7] = fitMeasures(fit, c("chisq","GFI", "AGFI", "RMSEA", "NFI", "NNFI", "CFI"))

var[i, 1:15] = parameterEstimates (fit, ci = FALSE, standardized = TRUE)[20:34, 4]

}

WARNING!!!

Error in fitMeasures(fit, c("chisq", "GFI", "AGFI", "RMSEA", "NFI", "NNFI",  : 

  lavaan ERROR: fit measures not available if model did not converge

[yrosseel]

On 05/18/2015 07:59 AM, Ainur Amira wrote:
>
> Hi everyone,
> I'm the beginner in R and lavaan, Can anyone help me give some suggestion?
> I try to run using a small sample size, (let say n=30), but the model
> only converges at n=55, but when I try using n=76, the model did not
> converge again. What should I do?

Well, either provide (much) better starting values, or get other/better
data. If your data is too far away from your model, it is not surprising
you get non-convergence, especially with small sample size.

Yves.

[Ainur Amira]

Thanks prof for your suggestion =)