Dear Naveed,
The situation can indeed be a bit confusing. The term
"observations" is actually used with two different senses in
SCALA. The first is in text such as the following:
Number of observations read : 62154
Number of unique reflections read : 5108
Number of observations output : 4922
Number of outliers rejected : 26
Number of observations rejected on Emax limit : 0
Number of observations outside resolution limits : 0
(observations outside resolution limits are omitted from the output file)
In this case, each part of a partially recorded reflection (those that
span several images) is counted as a separate observation, so for example
if a reflection spans 3 images it will have 3 observation.
However, in the summary at the end of SCALA it gives:
Overall InnerShell OuterShell
Total number of observations 23342 759 2493
Total number unique 4922 173 546
This is from the same logfile, but here the total number of observations
is AFTER the components of partials have been added together, so the
number is smaller. However the number of unique reflections is the same.
Typically, "number of observations" refers to this latter number rather
than the former. In a typical "Table 1", the "number of reflections" would
be the number of unique reflections (ie after merging symmetry mates)
while I feel that the "number of observations" should be reserved for the
total number of reflections measured (before merging symmetry mates). Very
often the multiplicity is also given in Table 1, which allows a rough
estimate of the total number of observations if this is not given
explicitly.
Note also that some reflections output by SCALA can be rejected at the
TRUNCATE step (those that are "too negative" and possibly E value
outliers, although the latter are normally rejected by SCALA). Thus the
number you see in the MTZ file as used by REFMAC may have fewer
reflections than those reported by SCALA as "unique" reflections.
Strictly speaking, the "Number of reflections" in Table 1 should be those
output by TRUNCATE (if used) rather than those reported in the SCALA
logfile, but I think many people use the value given by SCALA.
I hope this helps clarify things a bit.
Best wishes,
Andrew Leslie
> Dear CCP4 users,
>
> I am a bit confused about the use of these terms in regards to structure
> refinement statistics. When I process my data with SCALA, the program
> outputs statistics in terms of "total and unique numbers of observations".
> However, when I use the MTZ files with REFMAC, the final PDB file has
> "number of reflections". These values are of similar magnitude, but not
> identical. The issue is even more complicated when I look at tables of
> statistics between different journals. Authors often report unique
> reflections or unique observations.
>
> My questions are:
>
> 1) Are these two terms interchangable?
> 2) Are they relevant to the different stages of processing (e.g. data
> collection vs structure refinement)?
> 3) How do I rationalize the difference between the two values?
>
> I have read some of the widely used textbooks, but I am still confused
> when looking at publications. Any comments would be highly appreciated!
>
> Kind regards,
>
> Naveed Nadvi
>
> Faculty of Pharmacy,
> University of Sydney.
>
