Cluster data without knowing cutoff

Hi,

I have many datasets similar to this:

 0.51 0.60 0.45 1.20 0.55
 5.21 4.23 6.23 3.90 NaN
 0.70 0.68 0.54 0.22 NaN
 7.21 NaN  NaN  NaN  NaN

So the only structure within this data are the rows, containing multiple observations for an application. But obviously, rows one and three are similar and could be grouped together. Also row two and four are similar, because row 4 only contains one value and thus is "less heavy" than row 2.

I figured out, that this process is called clustering but it seems to need a cutoff. I cannot provide any information after which distance a new group starts, neither can I say how many groups exist.

I want Matlab finding possible groups in rows using an algorithm. I imagine, that someone else had this problem before so I'm asking if you know such an algorithm (as I just don't know what to search for).

Perhaps you can help me :)

[Walter Roberson]

It is not at all obvious to me that rows one and three are similar ?

When you say that the only structure is the rows, do you mean to imply that within each row, we could re-arrange the data arbitrarily -- since there is no structure to the columns ?