Getting Modification Indices for SEM model that uses bootstrap.

[Patrick Abi Nader]

Hi All,

I am using Stata 13 and utilizing structural equation modeling. Here is the code that I am using to estimate the relationship of a Latent variable Support with nine other variables. No problems so far. It may be irrelevant to include results for this first step but just in case they are included below. The 2nd step was to compute modification indices, but Stata lets me immediately know that modification indices cannot be computed from bootstrap results. I didn't know this, therefore is anyone out there more knowledgeable on this topic and able to point me to the right documentation to assist me in figuring out what I should be doing? I.E. I am trying to figure out if computing modification indices in my case is impossible because of some theoretical piece that I am missing, or if it's possible I just need to find the appropriate Stata language to do that?
Thank you for your time and feedback.
Patrick

**** Step 1, SEM model
sem (Support -> admin_barriers classroom_barriers expect_barriers ///
space othpa_materials student_beh class_env schedule bepa_materials), vce ///
(bootstrap, reps(100) seed(10)) method(mlmv)
****Step 2 Getting modification indices
estat mindices
modification indices not allowed with bootstrap results
r(322);

end of do-file

r(322);

Bootstrap replications (100)
----+--- 1 ---+--- 2 ---+--- 3 ---+--- 4 ---+--- 5 
..................................................    50
..................................................   100

Structural equation model                       Number of obs      =        83
Log likelihood = -661.93959                     Replications       =       100

 ( 1)  [admin_barriers]Support = 1
------------------------------------------------------------------------------------------
                         |   Observed   Bootstrap                         Normal-based
                         |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------------------+----------------------------------------------------------------
Measurement              |
  admin_barriers <-      |
                 Support |          1  (constrained)
                   _cons |   3.450451   .0659945    52.28   0.000     3.321104    3.579798
  -----------------------+----------------------------------------------------------------
  classroom_barriers <-  |
                 Support |   2.786461   1.741205     1.60   0.110    -.6262367     6.19916
                   _cons |   2.639186   .0881031    29.96   0.000     2.466507    2.811865
  -----------------------+----------------------------------------------------------------
  expect_barriers <-     |
                 Support |   1.716842   1.014308     1.69   0.091    -.2711646    3.704849
                   _cons |   2.131081   .0932142    22.86   0.000     1.948385    2.313778
  -----------------------+----------------------------------------------------------------
  space <-               |
                 Support |   2.669609   1.690747     1.58   0.114    -.6441948    5.983413
                   _cons |   2.515106   .0830723    30.28   0.000     2.352287    2.677925
  -----------------------+----------------------------------------------------------------
  othpa_materials <-     |
                 Support |   .9455596   .7652245     1.24   0.217     -.554253    2.445372
                   _cons |   2.436985   .0695391    35.04   0.000     2.300691    2.573279
  -----------------------+----------------------------------------------------------------
  student_beh <-         |
                 Support |   .6234749   .4137935     1.51   0.132    -.1875454    1.434495
                   _cons |   2.584307   .0714953    36.15   0.000     2.444179    2.724436
  -----------------------+----------------------------------------------------------------
  class_env <-           |
                 Support |   1.713544   1.196748     1.43   0.152    -.6320393    4.059127
                   _cons |   2.559527   .0663037    38.60   0.000     2.429574     2.68948
  -----------------------+----------------------------------------------------------------
  schedule <-            |
                 Support |   1.409653   .8098053     1.74   0.082    -.1775365    2.996842
                   _cons |    2.37505   .0746409    31.82   0.000     2.228757    2.521343
  -----------------------+----------------------------------------------------------------
  bepa_materials <-      |
                 Support |  -.0658193   .6584643    -0.10   0.920    -1.356385    1.224747
                   _cons |   2.746535   .0873328    31.45   0.000     2.575366    2.917704
-------------------------+----------------------------------------------------------------
    var(e.admin_barriers)|    .387718   .0709154                      .2709114     .554887
var(e.classroom_barriers)|   .1446005   .0415609                      .0823224    .2539927
   var(e.expect_barriers)|   .5229795     .10989                      .3464398     .789481
             var(e.space)|   .1066962   .0399623                       .051208    .2223107
   var(e.othpa_materials)|   .4079383   .0570093                      .3101982    .5364753
       var(e.student_beh)|   .4217189   .0692893                      .3056109    .5819388
         var(e.class_env)|   .2286193   .0789419                      .1161969    .4498124
          var(e.schedule)|   .3384308   .0721397                      .2228581    .5139387
    var(e.bepa_materials)|   .5066319   .0921325                      .3547309     .723579
             var(Support)|   .0644297   .0583223                      .0109287    .3798405
------------------------------------------------------------------------------------------

[Richard Williams]

You are trying to do a lot here. Only 83 cases and you are using bootstrapping and mlmv. How horrible would it be if you just did the MIs without doing the bootstrapping? The MIs are mostly a diagnostic device, I don't know that they have to be accurate to 6 decimal places.

[Patrick Abi Nader]

Thank you Richard! You are right I was trying to do a lot with a small sample. I resolved my challenges by not using bootstrap as it was an unnecessary step.
Patrick