# Difference between revisions of "R factor"

### From Online Dictionary of Crystallography

BrianMcMahon (talk | contribs) (Tidied translations and added German and Spanish (U. Mueller)) |
BrianMcMahon (talk | contribs) m (Tidied translations.) |
||

Line 1: | Line 1: | ||

− | <font color="blue">Facteur ''R''</font> (''Fr''). <font color="red">R-Faktor, Übereinstimmungsfaktor</font> (''Ge''). <font color="black">Fattore ''R''</font> (''It''). | + | <font color="blue">Facteur ''R''</font> (''Fr''). <font color="red">R-Faktor, Übereinstimmungsfaktor</font> (''Ge''). <font color="black">Fattore ''R''</font> (''It''). <font color="purple">R 因子</font> (''Ja''). <font color="brown">R-фактор</font> (''Ru''). <font color="green">Factor R</font> (''Sp''). |

## Latest revision as of 10:32, 15 December 2017

Facteur *R* (*Fr*). R-Faktor, Übereinstimmungsfaktor (*Ge*). Fattore *R* (*It*). R 因子 (*Ja*). R-фактор (*Ru*). Factor R (*Sp*).

## Definition

The term ** R factor** in crystallography is commonly taken to refer to the 'conventional'

*R*factor

[math]R = {{\sum | F_{obs} - F_{calc} | } \over {\sum |F_{obs} |}}[/math],

a measure of agreement between the amplitudes of the structure factors calculated from a crystallographic model and those from the original X-ray diffraction data. The *R* factor is calculated during each cycle of least-squares structure refinement to assess progress. The final *R* factor is one measure of model quality.

More generally, a variety of *R* factors may be determined to measure analogous residuals during least-squares optimization procedures. Where the refinement attempts to minimize the deviates of the squares of the structure factors (refinement based on [math]F^2[/math]), the *R* factor based on [math]F^2[/math] is used to monitor the progress of refinement:

[math]R(F^2) = {{\sum | F^2_{obs} - F^2_{calc} | } \over {\sum |F^2_{obs} |}}[/math].

Likewise, refinement based on *I* can be tracked using the Bragg *R* factor

[math]R_B = {{\sum | I_{obs} - I_{calc} | } \over {\sum |I_{obs} |}}[/math].

Even for refinement based on [math]F^2[/math] or *I*, the 'conventional' *R* factor may be calculated and quoted as a measure of model quality, in order to compare the resulting quality of models calculated at different times and with different refinement strategies.

The *R* factor is sometimes described as the **discrepancy index**.

*R* factor as a measure of structure quality

Theoretical values of *R* range from zero (perfect agreement of calculated and observed intensities) to about 0.6 for a set of measured intensities compared against a set of random intensities. *R* factors greater than 0.5 indicate very poor agreement between observed and calculated intensities, and many models with [math]R \ge 0.5[/math] will not respond to attempts at improvement. An early model with [math]R \le 0.4[/math] can usually be improved during refinement. A desirable target *R* factor for a protein model refined with data to 2.5 Å is considered to be [math]\sim 0.2[/math]. Small organic molecules commonly refine to [math]R \lt 0.05[/math]. However, the *R* factor must always be treated with caution, as an indicator of precision and not accuracy. Partially incorrect structures have been reported with *R* values below 0.1; many imprecise but essentially correct structures have been reported with higher *R* values.

## Weighted *R* factors

In practice, *weighted* *R* factors are more often used to track least-squares refinement, since the functions minimized are weighted according to estimates of the precision of the measured quantity *Y*:
[math]\sum w (Y_o - Y_c)^2[/math]
(*Y* being *F*, [math]F^2[/math] or *I*). The general term for a weighted residual is

[math]wR = \Big({{\sum |w |Y_o - Y_c|^2|}\over{\sum |wY^2_o}|}\Big)^{1/2}[/math].

The sum is usually computed over all reflections measured in the experiment. However, occasionally reflections are omitted from the calculation, either because they are believed to result from a systematic experimental error or are recorded with an intensity small compared with background noise. Any such selection may introduce statistical artefacts, and must always be described when reporting *R* factors.