Tue Jul 10 12:46:01 EDT 2012

Density of monetary transactions

See last post.  Compared to Sigma/Delta and sound waves, monetary
transactions seem to be different.  Why is this?

The problem is that simple smoothing doesn't seem to do the trick.  It
appears that some transactions have an intrinsic "bulkiness".  I.e. a
$1000 transaction that happens once a month (e.g. rent payment) is
different in kind than a $10 transaction that happens every day
(e.g. a meal at a fast food restaurant).

Modeling money streams using a density model seems to need to take
this into account: how much time does a transaction cover?  I.e. we
want to recover the "processing" that is done to the transaction.

Conclusion: monetary transactios are not "natural" in the sense that
they miss information, i.e. the intrinsic periodicity of a
transaction.  Is there a good way to attribute a time scale to
transactions that is meaningful?

The goal for (home) finance analysis is to distinguish recurring costs
(taxes, rent, food) from one-time costs.