The Definition of Connectivity about fimdbook HOT 3 CLOSED

All four patterns are connected.

But imagine that we wouldn't have column 1 in the data. Then "File matching" and "General" become unconnected. Thus, we for these patterns we say that column 1 connects columns 2 and 3.

Unconnected pattern may occur in practice when we combine data. These lead to estimation problems, so preferably one should connect the observed data by fostering overlap in the observations.

from fimdbook.

@stefvanbuuren

Thank you for your prompt answer!

To clarify,

• the term "observed data point" found in the book refers to the entire column (my 3-rd bullet in the question above) and,

• the term "connectivity" refers to row-wise overlapping of observed data between two columns located anywhere inside the dataset.

If we join all 4 panels above and no rows/columns are removed in the process, a barrier formed by columns 2 and 3 in "File Matching" would not allow columns 3 in "Univariate" and column 3 in "General" to connect because no "rook-like" movement can cross that barrier (similar to column 2 in "Monotone" and column 1 in "General" etc.).

A simple row-wise overlap however, will show these columns as connected.

Thank you!

from fimdbook.

OK, thanks. Now closing.

from fimdbook.