Discussion List Archives

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Some additional data items for the REFINE category

  • To: Multiple recipients of list <coredmg@iucr.org>
  • Subject: Re: Some additional data items for the REFINE category
  • From: Howard Flack <Howard.Flack@cryst.unige.ch>
  • Date: Mon, 21 May 2001 09:02:29 +0100 (BST)
> data_refine_ls_F_calc_accuracy
> data_refine_ls_F_calc_details

  It is not at all clear to me from the definitions that are given
whether the accuracy is meant to apply only to numerical approximations
(e.g. rounding) or whether the 'accuracy' also includes an estimation of
known deficiencies in the physical model expressed by
data_refine_ls_F_calc_formula.

> The weighted residual factors for all reflections used in the
> refinement, including explicitly the restraints applied in the
> least-squares process.

  Restraints are not necessarily reflections. How about:

The weighted residual factors for all observations used in the
least-squares refinement, including explicitly the intensities of
reflections and the restraints applied in the least-squares process.

> 
>                     sum|w[Y(obs)-Y(calc)]^2^|^1/2^
>                wR = ------------------------------  +
>                             sum|wY(obs)^2^|
> 
>                                   {sum~r~(w~r~|P(calc)-P(targ)|^2^)}^1/2^
>                                 + {--------------------------------}
>                                   {     sum~r~(w~r~P(targ)^2^)     }

  There are some funny things in there. In the first term, (a) I can not
see why you need absolute values of real terms which are weighted
squares (unless you have negative weights!) and (b) the square root is
definitely in the wrong place as it should apply to the quotient of the
sums in the divisor and the dividend (like term 2). In term (2) again I
do not see why there are absolute values of real quantities that are
positive definite.
  Is there any theory around that says one should calculate like this? I
mean specifically adding the square roots of the two quotients together,
rather than, for example, adding the two quotients and then taking the
square root, or even adding the two divisors, adding the two dividends,
forming the quotient and then taking the square root. 

Best wishes,
  H.

-- 
Howard Flack        http://www.unige.ch/crystal/ahdf/Howard.Flack.html
Laboratoire de Cristallographie               Phone: +41 22 702 62 49
24 quai Ernest-Ansermet             mailto:Howard.Flack@cryst.unige.ch
CH-1211 Geneva 4, Switzerland                   Fax: +41 22 702 61 08

[Send comment to list secretary]
[Reply to list (subscribers only)]