HMMTree: A computer program for hierarchical multinomial processing tree models

1. How to use this software: a brief user guide

Stahl, C. & Klauer, K.C. (2007). HMMTree: A computer program for hierarchical multinomial processing tree models. Behavior Research Methods, 39, 267- 273.

Klauer, K.C. (2006). Hierarchical multinomial processing tree models: A latent-class approach. Psychometrika, 71, 1 - 31.

1.1 Load a model and a data set

Model
1. Select File | Load equations or press CTRL + E. A file open dialog will show.
2. Pick an EQN file to load and select 'Open' or press ENTER.
3. The model will be loaded and displayed. Click on the 'Model' tab to verify that you selected the correct model.

Data
1. Select File | Load data or press CTRL + D. A file open dialog will appear.
2. At the bottom, select the file type (MDT or DAT) you wish to open.
3. Pick a data file to open and select 'Open' or press ENTER.
4. The data will be opened and displayed in a grid on the 'Data' tab, where you can verify your selection.

1.2 Compute model

Once you have successfully loaded a model and data file, the 'Run' button will be enabled.

1. Select the desired output options. By default, parameter estimates and confidence intervals are computed.
   • On the first run, parameter estimates need to be computed.
   • Confidence intervals are available when the Fisher information matrix is also computed. You can optionally select to print the inverse of the Fisher information.
   • Goodness-of-fit statistics for mean and variance/covariance structure are also available when the Fisher information is computed.
   • Also, the observed and expected means (classwise) and variances and covariances can be computed.
   • In case of more than one class, the posterior probability of a participant's class membership can be computed.
2. Click the 'Run' button or press ALT + R or F9 to start the computation. When computation is finished, the results are diplayed in the 'Output' tab.
3. Save the output to a text file with the File | Save output (or CTRL + S) or the File | Save output as... (or F12) commands.

• Additional computations (e.g., the Fisher information, observed and expected means and variances/covariances, and posterior probabilites) can be done without computing new parameter estimates.
• These computations will use the parameter values displayed in the 'Parameters' tab. They can be modified if desired.

The EQN Format

The EQN format used for this program is equivalent to the format used by AppleTree and MBT. An EQN file specifies the equations of a multinomial model.
The first line gives the number of equations in the model. Line 2 to line number of equations + 1 contain the model's equations.
An equation line contains three fields, separated by a space:

16
1 1 c1*r1
1 4 c1*(1-r1)
1 2 (1-c1)*u1*u1
1 3 (1-c1)*u1*(1-u1)
1 3 (1-c1)*(1-u1)*u1
1 4 (1-c1)*(1-u1)*(1-u1)
2 5 a1
2 6 (1-a1)
3 7 c2*r2
3 10 c2*(1-r2)
3 8 (1-c2)*u2*u2
3 9 (1-c2)*u2*(1-u2)
3 9 (1-c2)*(1-u2)*u2
3 10 (1-c2)*(1-u2)*(1-u2)
4 11 a2
4 12 (1-a2)

	equation ::= term | equation*term
	term ::= parameter | (1-parameter)
	

The MDT Format

The MDT data set consists of a data set description on the first line, followed by a line for each category, containing its number and its frequency, separated by a space. The end of a data set is marked by three equal signs. A file can contain multiple data sets.

An example of an MDT file with aggregated data:

Participants 1-3
1 8
2 1
3 7
4 13
5 7
===
	

Participant 1
1 4
2 0
3 1
4 5
5 2
===
Participant 2
1 3
2 0
3 4
4 2
5 3
===
Participant 3
1 1
2 1
3 2
4 6
5 2
===
	

The DAT Format (tab-separated matrix)

The DAT format allows you to easily use data from other statistics software for hierarchical multinomial model estimation. It basically consists of a matrix of individuals' category frequencies.

The first line of a DAT file is interpreted as data set description. It can contain labels for each category or, alternatively, a description of the data set.

Each additional line is interpreted as an individual's set of category frequencies. Category frequencies are read from left to right. The first value is assigned to category 1, the second value is assigned to category 2, etc. Values on a line have to be separated by (one or multiple) TAB(s) or space(s).

A DAT file may contain individuals' numbers or codes as the first value of each line. if the second line of a DAT file contains one more value (delimited by tabs or spaces) than the model contains categories, the first value in each line is interpreted as a participant code or number.

An example with category labels:

A	B	C	D	E
4	0	1	5	2
3	0	4	2	3
1	1	2	6	2 
	

Subjects 1-3:
4	0	1	5	2
3	0	4	2	3
1	1	2	6	2 
	

A	B	C	D	E
S1	4	0	1	5	2
S2	3	0	4	2	3
S3	1	1	2	6	2 
	

SAVE TRANSLATE OUTFILE='filename.dat' /TYPE=TAB /MAP /REPLACE /FIELDNAMES.

2. Contact the authors

Christoph Stahl & Karl Christoph Klauer
Institute for Psychology
Albert-Ludwigs-Universitaet Freiburg
D-79085 Freiburg i. Br.
Germany

E-Mail: stahl@psychologie.uni-freiburg.de