Rows are weighted.
Place the weights of each row in the first numerical column, just after the row identifiers.

Original values

Mean center Columns

The graphical representation is based on column-centered data: column mean is subtracted from each value, so that mean value of each column is 0.

Mean center Rows

Normalize Columns (Z-score)

Normalize Rows (Z-score)

Row-column interactions

The graphical representation is based on data after subtraction of row and column deviation.
The sums of each row and each column are equal to zero.
These values correspond to the interaction term issued from a two-factor analysis of variance (rows x columns). This term can be expressed as:

Same Color Levels

Specific Color Levels

The graphical representation is based on column-centered data: column mean is subtracted from each value, so that mean value of each column is 0.

1 - Numerical Data


2 - Binary Data


3 - Nominal (categorical) Data

Distances or Index of dissimilarity used for clustering and seriation

Euclidean distance

Manhattan distance

Chebyshev distance

Pearson distance

Clustering, and seriation methods, are based on Pearson disimilarity s =1-r, where r is the usual Pearson correlation coefficient. This metric varie between 0 and 2.
s = 0 when two curves have "identical" shape, but different magnitude.
s = 1 when curves variations are completely independent.
s = 2 when two curves have opposite shape.

Squared Pearson distance

Clustering, and seriation methods, are based on Pearson disimilarity , where r is the usual Pearson correlation coefficient. This metric varies between 0 and 1.
s = 0 when two curves have "identical" or "opposite" patterns, but different magnitude.
s = 1 when curves variations are completely independent.
This disimilarity will be useful for very specific situation. It will be difficult to read clustering and seriation results. Two very different colour patterns could be put together.

Jaccard's distance

Dice's dissimilarity indice

Read dissimilarity values from a file

If you want a special distance, or dissimilarity, you can compute your own distance or dissimilarity values with another application and import these values in PermutMatrix from a file. This file must be a triangular matrix in a standard text file format (ASCII) as for data file format.
In this case, you have to provide two files. One for distances, or dissimilarities, between rows and one for distances, or dissimilarities, between columns.

aggregation criteria

Complete linkage

The distance between two classes A and B is the maximum of all pairwise distances between items contained in A and B.

Single linkage

The distance between two classes A and B is the minimum of all pairwise distances between items contained in A and B.

McQuitty 's method (WPGMA)

The distance between two classes A and B is an weighted mean of all pairwise distances between items contained in A and B. The distance of a group of two classes A and B to a class C is the unweighted average of the distances between A and C on one hand, and B and C on the other.

Average linkage (UPGMA)

The distance between two classes A and B is the unweighted mean of all pairwise distances between items contained in A and B. So, the distance of a group of two classes A and B to a class C is the average of the distances between A and C on one hand, and B and C on the other hand, the average being weighted by the size of each class.

Ward's minimum variance

Seriation Criteria:

Multiple-fragment heuristic (MF)

Unidimentional Scaling (UDS)

Inertia around diagonal

Bipolarization

Minimum path length (TSP)