Using Categorical (Text) Data in SimDec

Recommended Encoding Method

Encode categories based on their average output value. This will ensure nicely ordered visual decomposition.

The rule is simple:

Categories with lower average output get lower numbers.
Categories with higher average output get higher numbers.

Step-by-Step Encoding

Assume:

• a categorical input X
• an output variable Y

1. Compute Mean Output by Category

Calculate the mean of Y for each category of X.

2. Sort Categories by Mean Output

Sort categories from lowest to highest mean.

3. Assign Numeric Codes

Assign integers starting from 1.

4. Replace Text with Numbers

Replace all text values in the dataset with the assigned codes.

‍

Notes

• Use the output of interest when computing means.
• If there are multiple outputs, encode using the primary one.
• Recompute the encoding if the simulation model changes.

Summary

To use categorical variables in SimDec:

1. Compute mean output by category
2. Sort categories by mean
3. Assign numeric ranks
4. Replace text values

Code examples

Python

import pandas as pd

# df columns: 'run', 'X', 'Y'

# Example: df = pd.read_csv("sim.csv")

# 1) mean output per category

means = df.groupby("X")["Y"].mean()

# 2) sort categories by increasing mean

ordered = means.sort_values().index

# 3) assign codes 1..K

code_map = {cat: i + 1 for i, cat in enumerate(ordered)}

# 4) replace X with numeric codes

df["X_encoded"] = df["X"].map(code_map)

# Optional: keep only numeric input if you want

# df = df.drop(columns=["X"])

print(code_map)

print(df.head())

‍

R

# df columns: run, X, Y

# df <- read.csv("sim.csv", stringsAsFactors = FALSE)

# 1) mean output per category

m <- tapply(df$Y, df$X, mean)

# 2) sort categories by increasing mean

cats <- names(sort(m))

# 3) assign codes 1..K

code_map <- setNames(seq_along(cats), cats)

# 4) replace X with numeric codes

df$X_encoded <- unname(code_map[df$X])

print(code_map)

head(df)

‍

MATLAB

% T is a table with variables: run, X (string/cellstr/categorical), Y (double)

% Example: T = readtable("sim.csv");

% Ensure X is categorical for grouping

T.X = categorical(T.X);

% 1) mean output per category

[G, cats] = findgroups(T.X);

meanY = splitapply(@mean, T.Y, G);

% 2) sort categories by increasing mean

[~, idx] = sort(meanY, 'ascend');

cats_sorted = cats(idx);

% 3) assign codes 1..K

codes = (1:numel(cats_sorted))';

% 4) replace X with numeric codes

% Build mapping via ismember

[tf, loc] = ismember(T.X, cats_sorted);

T.X_encoded = codes(loc);

% Optional: show mapping

mapping = table(cats_sorted, codes, 'VariableNames', {'Category','Code'});

disp(mapping);

head(T)

‍

Excel

Assume your data is in an Excel Table named Sim with columns:

• Sim[Simulation run]
• Sim[X] (text categories)
• Sim[Y] (output)

Step A — Create a mapping table (Category → MeanY → Code)

1. List unique categories (e.g., in H2):

=SORT(UNIQUE(Sim[X]))


2. Mean output for each category (in I2, next to the spilled list):

=AVERAGEIF(Sim[X], H2#, Sim[Y])


3. Sort categories by mean (create sorted category list, in K2):

=SORTBY(H2#, I2#, 1)


4. Create codes 1..K (in L2):

=SEQUENCE(ROWS(K2#))


Now you have a mapping table:

• K2# = categories sorted by mean Y
• L2# = numeric codes

Step B — Replace X with code (X_encoded)

Add a new column Sim[X_encoded] and use:

=XLOOKUP([@X], $K$2#, $L$2#)


That’s it: Sim[X_encoded] is your numeric version of X.

‍