when in output there are two result a robust for CFI,TLI and RMSEA

2,474 views
Skip to first unread message

Parisa Ganje

unread,
Jun 6, 2021, 8:13:41 AM6/6/21
to lavaan
Dear All,
I would appreciate it if you guide me regarding my model and following questions:
I have RUN  a reciprocal model:
'
 Y2 ~ Y1+Y0+ X1
  Y1 ~ Y0+ X0
 X2 ~ X1+ X0+Y1
  PA1 ~ X0+Y0
  Y0 ~~ X0
  Y1 ~~ X1
  Y2 ~~ X2'                       
ReciprocalModel.fit<- sem(ReciprocalModel,data = X22May,se = "robust",missing="pairwise",ordered=c("X0","X1","X2"))
                                              
summary(ReciprocalModel.fit, fit.measures = TRUE,standardized = TRUE,ci = TRUE,   rsquare = TRUE)

In out put I got two robust result for CFI,TLI and RMSEA that one of them is NA.
1)which values should be considered for fit indices? values highlighted in green are robust. values highlighted in yellow are robust but the result is NA.
output:

lavaan 0.6-8 ended normally after 98 iterations

  Estimator                                       DWLS

  Optimization method                           NLMINB

  Number of model parameters                        25

                                                                         Used       Total

Number of observations                         17452       17640

  Number of missing patterns                        46                                                                         

Model Test User Model:

                                              Standard      Robust

  Test Statistic                            2.118        2.797

  Degrees of freedom                      2           2

  P-value (Chi-square)               0.347       0.247

  Scaling correction factor                                  0.757

  Shift parameter                                            0.001

       simple second-order correction

Model Test Baseline Model:

  Test statistic                              5830.631    5214.212

  Degrees of freedom                       15             15

  P-value                                        0.000            0.000

  Scaling correction factor                                  1.119   

User Model versus Baseline Model:

  Comparative Fit Index (CFI)                1.000       1.000

  Tucker-Lewis Index (TLI)                    1.000        0.99                                                        

  Robust Comparative Fit Index (CFI)                       NA

  Robust Tucker-Lewis Index (TLI)                               NA

Root Mean Square Error of Approximation:

  RMSEA                                                0.002                          0.005

  90 Percent confidence interval - lower         0.000       0.000

  90 Percent confidence interval - upper         0.015       0.017

  P-value RMSEA <= 0.05                          1.000       1.000                                                         

  Robust RMSEA                                                  NA

  90 Percent confidence interval - lower                     0.000

  90 Percent confidence interval - upper                        NA

 

Standardized Root Mean Square Residual:

  SRMR                                           0.014       0.014

Parameter Estimates:

  Standard errors                           Robust.sem

  Information                                 Expected

  Information saturated (h1) model        Unstructured               

Terrence Jorgensen

unread,
Jun 13, 2021, 8:55:48 AM6/13/21
to lavaan
You can search for previous responses to the same question posted multiple times previously on this forum.  When the ones called "robust" are available, use them; when they are NA, use the standard formulas that replace the standard chi-squared with the robust ("scaled") chi-squared, which are in the "Robust" column but in the rows next to the standard RMSEA/CFI/TLI.

Terrence D. Jorgensen
Assistant Professor, Methods and Statistics
Research Institute for Child Development and Education, the University of Amsterdam

Sofia

unread,
May 6, 2022, 9:23:55 AM5/6/22
to lavaan
Dear Terrence,

Apologies if the answer to this question has been given but I just cannot find it. 
What should I cite in the instance in which Robust fit indices are set to NA and I should use the standard formulas? (as bove) Would it be the same Brosseau-Liard papers that justify the use of robust indices?

Thank you!

Terrence Jorgensen

unread,
May 6, 2022, 1:31:05 PM5/6/22
to lavaan
What should I cite in the instance in which Robust fit indices are set to NA and I should use the standard formulas? (as bove) Would it be the same Brosseau-Liard papers that justify the use of robust indices?

Sure, they also discuss the standard formulas to explain why they are suboptimal.  Savalei has published a couple more papers on adjustments for the cases when lavaan returns NAs:


Reply all
Reply to author
Forward
0 new messages