Chains or strings are defined with absolute precision, but groups are very loosely defined here.
I think of a group as a collection of chains of the same colour that could be connected regardless of what the opponent did.
There are obvious problems with this definition. For loose groups, early in a game, it will often be unclear if connection could be forced.
For groups that are almost connected (by a diagonal connection for example) disconnection might be achieved if the connection was attacked as a ko threat which where the ko was more valuable than the connection. Still it is reasonable to ignore this in the definition. There are other circumstances in which there might simply be more important things to do than respond to attempts to disconnect a group.
In many circumstances the player would never wish to actually make a group solidly connected (turning it into a chain). For example it might destroy an eye, or simply be a waste of territory (in Japanese rules). This doesn't really matter to the definition.
Your definition sounds good. However, I find that actual usage of the word "group" seems vaguer than your definition. Quite often life and death problems say "save the group" but the group in question also involves loose stones that can easily be (and should be) sacrificed. How's that?
"Group" is not a precise term. It is, like many, perhaps most, words, a fuzzy term. I think that language is fuzzy for good reason, and I do not see a compelling reason to make "group" precise. "Group" is a Gestalt. It depends upon the human capacity to recognize patterns and perceive wholes.
[Harry_Fearnley] This precision is more apparent than real. What about hanezeki -- see http://senseis.xmp.net/?HaneSeki ? How many groups are there? ... arguably as many as there are chains!