潜变量混合增长模型-经管之家官网！

各位大神好，我在用MPLUS处理追踪数据的时候遇到一些困难，想问问如何解决
有三次数据，采用多重插补法处理缺失值，然后建立了潜变量混合增长模型，然后死循环了，下面是MPLUS代码
TITLE: Missing Data;
DATA: FILE IS 3depress.dat;
VARIABLE:
NAME are gender f1-f18 b1-b18 c1-c18;
USEVARIABLES =gender f1-f18 b1-b18 c1-c18;
MISSING are ALL (9999);
DATA IMPUTATION:
IMPUTE = b1-b18 c1-c18;
NDATASETS = 50;
save = depressimp*.dat;
ANALYSIS:
type = basic;
bseed = 79566;
chains = 1;
TITLE: Summary Statistics with Multiply Imputed Data;
DATA: FILE IS depressimplist.dat;
TYPE = imputation;
VARIABLE:
NAME are gender f1-f18 b1-b18 c1-c18;!三次追踪题项
USEVARIABLES = gender f1-f18 b1-b18 c1-c18 DFOUR DFIVE DSIX;
class=c(2);
ANALYSIS:
ESTIMATOR = ML;
INFORMATION = observed;
TYPE=MIXTURE;
START=500 50;
processor=4;
DEFINE:
DFOUR=MEAN(f1-f18);!第一次均值
DFIVE=MEAN(b1-b18);!第二次均值
DSIX=MEAN(c1-c18);!第三次均值
MODEL:
%OVERALL%
i s|DFOUR@0 DFIVE@1 DSIX@2;
OUTPUT:TECH11 TECH14 TECH4 ;
SAVEDATA:FILE=DEPRESS2.dat;
SAVE=cprob;
plot:TYPE is PLOT3;
直接删除缺失值跑数据，提示矩阵非正定，已经尝试过将斜率方差固定为0，但是得到的分组人数差距太大，一组400多，一组10，我想问问是不是模型不成立，或者数据有问题
WARNING:THE LATENT VARIABLE COVARIANCE MATRIX (PSI)IN CLASS 1IS NOT POSITIVE DEFINITE.THIS COULD INDICATE A NEGATIVE VARIANCE/RESIDUAL VARIANCE FOR A LATENT VARIABLE, A CORRELATION GREATER OR EQUAL
TO ONE BETWEEN TWO LATENT VARIABLES, OR A LINEAR DEPENDENCY AMONG MORETHAN TWO LATENT VARIABLES.CHECK THE TECH4 OUTPUT FOR MORE INFORMATION.PROBLEM INVOLVING VARIABLE S.
这是TECH4输出的结果
ESTIMATED MEANS FOR THE LATENT VARIABLES
I S
________ ________
1 1.645 0.008
ESTIMATED COVARIANCE MATRIX FOR THE LATENT VARIABLES
I S
________ ________
I 0.003
S 0.009 -0.005
ESTIMATED CORRELATION MATRIX FOR THE LATENT VARIABLES
I S
________ ________
I 1.000
S 999.000 999.000