Multinomial & Conditional Logit Regressions

Estimate discrete-choice models using the mlogit R package — with no R install required
Long- or wide-format CSV upload, with automatic reshape for wide
Filter to a subset of rows before estimating
Add variables with alternative-invariant or alternative-specific slopes
View coefficients, fit statistics, and ready-to-paste prediction equations
Compute predicted choice probabilities for any input scenario

Disclaimer

Results may contain errors. Output may be inaccurate or incomplete. Verify all results independently.
No warranty. Provided "as is" without any warranty.
Privacy: Data never leaves your browser — everything runs locally in WebAssembly.

📧 Report issues: Email Dr. Cosguner at kcosgun@iu.edu.

1. Choose Data Format

Pick how your data is laid out before uploading. The required columns and the data preview both depend on this choice.

Long format (one row per alt per occasion) Wide format (one row per occasion)

Long Format recommended

Each row represents one alternative within one choice occasion. If there are 3 alternatives per choice, each occasion has 3 rows. Required columns: a Choice ID, an Alternative name, a binary Choice indicator, and any alternative-specific attributes (e.g., Price).

ChoiceID  Alternative  Price  Chosen
1         Coke         0.64   0
1         Pepsi        0.85   1
1         NoPurchase   0      0
2         Coke         0.87   0
2         Pepsi        0.63   0
2         NoPurchase   0      1
3         Coke         0.71   1
3         Pepsi        0.79   0
3         NoPurchase   0      0
4         Coke         0.83   0
4         Pepsi        0.69   1
4         NoPurchase   0      0

↕ scroll · ↔ scroll

Wide Format

Each row represents one choice occasion. There is a Choice ID column, one binary choice column per alternative (including the outside option), plus alternative-specific attribute columns named with a per-alternative suffix (e.g., PriceCoke, PricePepsi).

ChoiceID  PriceCoke  PricePepsi  PriceNoPurchase  ChoiceCoke  ChoicePepsi  ChoiceNoPurchase
1         0.64       0.85        0                0           1            0
2         0.87       0.63        0                0           0            1
3         0.71       0.79        0                1           0            0
4         0.83       0.69        0                0           1            0

↕ scroll · ↔ scroll

The tool automatically reshapes wide data to long before estimation.

2. Upload Your Data

Upload a CSV in long format (one row per alternative per choice occasion).

📊

Click to upload or drag and drop a CSV

Long format · CSV only

Data Preview (first 8 rows)

4. Configure Choice Data

Tell mlogit.data() how the columns map to the choice structure.

Choice ID Variable (groups rows per occasion)

Alternative Variable (names the options)

Choice Variable (0/1 indicator)

Reference Alternative (baseline)

5. Specify the Model

Independent variables are split into two groups based on whether they vary by alternative within a choice occasion. Choice-Specific IVs (e.g., Price) take a different value across alternatives within an occasion — they can use either a single slope across alternatives or a per-alternative slope. Choice-Invariant IVs (e.g., Summer, Income) are constant across alternatives within an occasion — they must use a per-alternative slope, because a single slope would cancel out of the choice probabilities.

Choice-Specific Independent Variables (value varies by alternative — e.g., Price)

No variables available yet.

Choice-Invariant Independent Variables (value constant within an occasion — e.g., Summer)

No variables available yet.

Intercept

Alternative-Specific Intercepts No Intercept

6. Results

Coefficient Estimates

Select the table and copy (Ctrl+C / Cmd+C) to paste into Excel.

Decimals:

Model Fit

Prediction Equations

Include coefficients with:

Any p-value p < 0.10 p < 0.05 p < 0.01

Full R Output

Make a Prediction

Enter values for each non-baseline alternative below to compute predicted choice probabilities (the baseline alternative has utility 0 by definition).

Predicted choice probabilities