[fitsbits] CRPIX clarification

[Mark Calabretta]

On Mon 2008/06/16 17:36:27 MST, Rob Seaman wrote
in a message to: Mark Calabretta <mcalabre at atnf.csiro.au>
and copied to: Eric Greisen <egreisen at nrao.edu>, FITSbits <fitsbits at nrao.edu>,
      William Pence <William.Pence at nasa.gov>

Dear Rob,

>The question is whether this discussion of randomness adds anything to  
>the specific paragraph.

I believe it is useful to explain one thing by pointing out the
distinction with another.  If the reader doesn't know anything about
random groups data then they can read over it.  (That's why it's in
parentheses.)

BTW, "random" is not an apt description, "irregular" would have been
better.

>Isn't the distinction here between a  
>repeating lattice and an individual datum, not between regular and  
>random structure?

The distinction is between defining a rule for computing the world
coordinates of each datum, and recording the world coordinates of
each datum.

In the former case, neither pixel coordinates nor world coordinates are
recorded.  Pixel coordinates are inferred from array indices and that
is only possible because the data form a regular lattice.

That is not possible for the random parameter component of random groups
data because they are not regular.  Therefore, you have to record the
world coordinate of each datum, or more usually, each array of data.
That is what the random parameters themselves are.

Regards, Mark