A. Medical Diagnosis Example from Chapter 3 - Unrestricted Two Class Model

1234567890123456789012345678901234567890123456 (Columns - not part of input}

Pleural thickening - unrestricted 2 class  {Title line}
 3 2  1692  2000         1 1                 1 {See a., below}
 2 2 2         {Levels of 3 manifest variables}
(8F4.0)       {Variable format for frequencies}
15130021005900110023001900120034   {See b., below}
0.30    0.70                          {LC proportions-initial estimates}
0.4     0.6     0.6     0.4        {See c., below}
0.4     0.6     0.6     0.4
0.4     0.6     0.6     0.4

 a. Problem line: Columns 1-2 = number of manifest variables (number of levels per variable on next line in two-digit fields); Columns 3-4 = number of latent classes; Columns 5-10 = sample size; Columns 11-16 = maximum iterations allowed (default = 500); Columns 17-24 = convergence criterion (blank means default = .00005); Column 26 = 1 results in fitting model of complete independence; Column 28 = 1 results in assignment of response vectors to latent classes; Column 46 = 1 results in assessment of rank of asymptotic covariance matrix.
 b. Frequencies for observed response vectors in four-digit fields in accordance with preceding variable format line; response vectors are ordered with leftmost digits varying more slowly - 000, 100, 010, 110, 001, 101, 011, 111.
 c. Initial estimates for latent class proportions with each row representing one manifest variable and each pair of columns representing one latent class; present example (in eight-digit fields) is:
p_l1^AX ; 1 - p_1l^AX  ; p _l2^AX  ; 1 - p₁₂^AX
p_1l^BX ; 1 - p_1l^BX   ; p _l2^BX  ; 1 - p₁₂^BX
p_1l^CX ; 1 - p_1l^CX   ; p _l2^CX  ; 1 - p_l2^CX

B. Medical Diagnosis Example from Chapter 3 - Restricted Two Class Model (V)

1234567890123456789012345678901234567890123456 (Columns - not part of input}

Pleural thickening data - Model V
 3 2  1692  2000         1 1         1       1 {See a., below}
 2 2 2
(8F4.0)
15130021005900110023001900120034
0.50    0.50
0.4     0.6     0.6     0.4
0.4     0.6     0.6     0.4
0.4     0.6     0.6     0.4
   4   0   2   0      {See b., below}
   4   0   3   0
   4   0   2   0

 a. Column 38 = 1 indicates restrictions on conditional probabilities.
 b. Restrictions on conditional probabilities in same relative fields as initial estimates (note c. in A., above). Equal integers imply equality restrictions (0 indicates value constrained by summation restrictions such as p_l2^AX = 1 - p₁₁^BX  ). The indicated restrictions are: p_l1^AX =  p₁₁^BX  = p ₁₁^CX  and p₁₂^AX  =  p ₁₂^CX (note that

p ₁₂^BX is unconstrained).

C. Left-Right Clinical Scale Example from Chapter 4 - Linear Scale, Error Models

1234567890123456789012345678901234567890123456 (Columns - not part of input}

Whitehouse Left-Right Data - Linear Scale, IO Model
 3 4   573  2000         1 1         1       1
 2 2 2
(8F3.0)
170073006254000001000069
0.2     0.2     0.3     0.3
0.8     0.2     0.8     0.2     0.2     0.8     0.2     0.8
0.8     0.2     0.8     0.2     0.2     0.8     0.2     0.8
0.8     0.2     0.8     0.2     0.2     0.8     0.2     0.8
   2   0   0   4   0   4   0   4     {See a., below}
   2   0   2   0   0   4   0   4
   2   0   2   0   2   0   0   4
 a. These restrictions are for the intrusion-omission error model and correspond to equation (4.2). For the Proctor model, with the restrictions in equation (4.1), these lines would be:

   2   0   0   2   0   2   0   2
   2   0   2   0   0   2   0   2
   2   0   2   0   2   0   0   2

For a variable-specific error model, with the restrictions in equation (4.3), these lines would be:

   2   0   0   2   0   2   0   2
   4   0   4   0   0   4   0   4
   6   0   6   0   6   0   0   6

And for a latent-class-specific error model with the restrictions in equation (4.4), these lines would be:

   2   0   0   4   0   6   0   8
   2   0   4   0   0   6   0   8
   2   0   4   0   6   0   0   8

D. Stouffer-Toby Example from Chapter 4 - Latent Distance Model

1234567890123456789012345678901234567890123456 (Columns - not part of input}

Stouffer-Toby data - Latent Distance Model
 4 5  216.               1 1         1       1
 2 2 2 2
(16F3.0)
 20 38  6 25  9 24  4 23  2  7  1  6  2  6  1 42
0.3     0.1     0.2     0.2     0.2
0.8     0.2     0.8     0.2     0.8     0.2     0.8     0.2     0.8     0.2
0.8     0.2     0.8     0.2     0.8     0.2     0.8     0.2     0.8     0.2
0.8     0.2     0.8     0.2     0.8     0.2     0.8     0.2     0.8     0.2
0.8     0.2     0.8     0.2     0.8     0.2     0.8     0.2     0.8     0.2
   2   0   0   2   0   2   0   2   0   2        {See a., below}
   3   0   3   0   0   6   0   6   0   6
   4   0   4   0   4   0   0   7   0   7
   5   0   5   0   5   0   5   0   0   5
 a. These restrictions are for the latent distance model based on the constraints in equation 4.7 and in the paragraph below equation 4.7.

E. Two-Group Cheating Example from Chapter 6 - Model of Complete Homogeneity

1234567890123456789012345678901234567890123456 (Columns - not part of input}

DQ Data Set - Model of Complete Homogeneity
 5 2   317               1 1         1      1    {see a., below}
 2 2 2 2 2
(16F3.0)
 99  5  1  3  1  1  0  1 18  1  1  2  2  1  1  0  {see b., below}
107  5 11  8  6  0  1  0 28  2  3  2  3  1  1  2
0.15    0.85
0.3     0.7     0.7     0.3
0.3     0.7     0.7     0.3
0.3     0.7     0.7     0.3
0.3     0.7     0.7     0.3
.42318  .56782  .42318  .56782     {see c., below}
   2   0   6   0
   3   0   7   0
   4   0   8   0
   5   0   9   0
   1   0   1   0        {see d., below}

 a. Number of variables is five - four cheating items and sex of student.
 b. Frequencies for 16 response vectors for males followed by frequencies for females.
 c. Start values for proportions of males (.42318) and females (.56782) are set at the same values for the two latent classes. These are the actual proportions of males and females in the sample.
 d. Final values for proportions of males and females are fixed at start values.