用一组数据在Mplus中做LCA分析，遇到了如下的显示，求助这是什么意思？应该如何解决？

RANDOM STARTS RESULTS RANKED FROM THE BEST TO THE WORST LOGLIKELIHOOD VALUES

9 perturbed starting value run(s) did not converge in the initial stage
optimizations.

Final stage loglikelihood values at local maxima, seeds, and initial stage start numbers:

           -2066.028  928287           197
           -2066.028  311214           64
           -2066.028  321390           133
           -2066.028  618000           190
           -2066.028  unperturbed      0
           -2768.522  509733           130
           -2768.522  329127           185
           -3179.751  813779           92
           -3179.751  802779           122
           -3196.877  347515           24
           -3223.059  432148           30
           -3223.059  761633           50
           -3223.059  313407           132
           -3341.338  830392           35

36 perturbed starting value run(s) did not converge.



THE BEST LOGLIKELIHOOD VALUE HAS BEEN REPLICATED.  RERUN WITH AT LEAST TWICE THE
RANDOM STARTS TO CHECK THAT THE BEST LOGLIKELIHOOD IS STILL OBTAINED AND REPLICATED.

     THE STANDARD ERRORS OF THE MODEL PARAMETER ESTIMATES MAY NOT BE
     TRUSTWORTHY FOR SOME PARAMETERS DUE TO A NON-POSITIVE DEFINITE
     FIRST-ORDER DERIVATIVE PRODUCT MATRIX.  THIS MAY BE DUE TO THE STARTING
     VALUES BUT MAY ALSO BE AN INDICATION OF MODEL NONIDENTIFICATION.  THE
     CONDITION NUMBER IS      -0.983D-22.  PROBLEM INVOLVING THE FOLLOWING PARAMETER:
     Parameter 15, %C#2%: [ NEG2 ]


THE MODEL ESTIMATION TERMINATED NORMALLY