LPA Latent Class Means Extractor

Extract per-class variable Means and Class Counts from Mplus LPA output files and generate a multi-sheet Excel workbook (one sheet per model).

When to Use

User has multiple Mplus .out files from LPA models (e.g., 2-class through 6-class)
User wants a table showing, for each latent class, the Means of each analysis variable
User also wants Class Counts (integer counts) in the same table
Output format: one Excel sheet per model, with Class1/Class2/... as columns and variables as rows

Workflow

Step 1: Identify inputs

Ask the user or infer from context:

Which directory contains the .out files?
What is the file naming pattern? (e.g., T3 LPA*.out)
Which variables to extract as row labels? (e.g., T3PF_hos,T3PF_con,T3PF_dis,T3PM_hos,T3PM_con,T3PM_dis,T3CF_m)
What prefix to strip from filenames for sheet labels? (e.g., T3 to get LPA2, LPA3)
Output Excel filename?

Step 2: Run the extraction script

python scripts/extract_lpa_means.py <directory> \
    --pattern "GLOB_PATTERN" \
    --output OUTPUT.xlsx \
    --variables "VAR1,VAR2,VAR3,..." \
    [--sheet-prefix LPA] \
    [--row-prefix "PREFIX_TO_STRIP"]

Arguments:

directory (required): Path to the directory containing .out files
--pattern (default *.out): Glob pattern to match files
--output (default LPA_Class_Means.xlsx): Output Excel filename
--variables (required): Comma-separated variable names to extract as row labels
--sheet-prefix (default LPA): Prefix for sheet names (e.g., LPA -> LPA2, LPA3)
--row-prefix (default empty): Prefix to strip from filenames for sheet labels. E.g., if files are T3 LPA2.out and you want sheet label LPA2, use --row-prefix "T3 "
--row-suffix (default .out): Suffix to strip from filenames for sheet labels

Step 3: Present the result

After the script completes, call present_files with the output Excel file path. Also summarize the extracted values in a markdown table in the reply.

Extracted Values

Each sheet in the Excel workbook has the following structure:

| Row | Content | Source in .out file | |-----|---------|-------------------| | Header | "", Class1, Class2, ... | Generated | | Class Counts | Integer count per class | Class Counts and Proportions section (under "BASED ON THEIR MOST LIKELY LATENT CLASS MEMBERSHIP") -> 2nd column | | Variable rows | Mean estimate per class | MODEL RESULTS -> Latent Class N -> Means sub-section -> Estimate column (1st numeric column) |

Extraction details

Class Counts: Extracted from the "Class Counts and Proportions" section that appears under "BASED ON THEIR MOST LIKELY LATENT CLASS MEMBERSHIP". The 2nd column (integer counts) is used.
Variable Means: Extracted from the MODEL RESULTS section. For each Latent Class N block, the Means sub-section is parsed. Each variable line has the format VAR_NAME estimate SE est/SE p-value; the estimate (1st numeric column) is extracted.
Variable matching: Variable names are matched case-insensitively (uppercased). If a variable in the user's list is not found in the .out file (e.g., not in USEVARIABLES), its cell is left empty.
Number of classes: Inferred from the number of Class Counts entries, or from the highest Latent Class N in MODEL RESULTS, or from the numeric suffix in the filename.

Notes

The script requires openpyxl. If not installed, install it first with pip.
Mean values are formatted with 0.000 (3 decimal places) in Excel.
Class Counts are displayed as integers.
The Excel output uses Times New Roman font, centered alignment, thin borders, and light blue header fill.
Files with spaces in names (e.g., T3 LPA2.out) are supported.
If a variable is not in the model (not in USEVARIABLES), its row will be empty -- this is expected behavior, not an error.
The MODEL RESULTS section is isolated by finding MODEL RESULTS at the start and FINAL CLASS COUNTS / QUALITY OF CLASSIFICATION / TECHNICAL at the end to avoid cross-section contamination.
Within each Latent Class N block, only the Means sub-section (before Variances) is parsed.

Example

python scripts/extract_lpa_means.py /path/to/T3 \
    --pattern "T3 LPA*.out" \
    --output T3_LPA_Class_Means.xlsx \
    --variables "T3PF_hos,T3PF_con,T3PF_dis,T3PM_hos,T3PM_con,T3PM_dis,T3CF_m" \
    --sheet-prefix LPA \
    --row-prefix "T3 "

This produces a 5-sheet Excel file:

Sheet LPA2: 2 class columns (Class1, Class2)
Sheet LPA3: 3 class columns (Class1, Class2, Class3)
Sheet LPA4: 4 class columns
Sheet LPA5: 5 class columns
Sheet LPA6: 6 class columns

Each sheet has rows: Class Counts, then one row per variable.