Reference Material Plotting Template
Version 2-1.2 May 2010
Welcome to the Rocklabs Template for plotting the results from the analysis of your Reference Materials.
This template may be used for the plotting of Reference Material results for any element, however the comments associated
with the statistical analysis on the "Enter Data" sheet and the "Explanation of Terms" sheet are intended primarily for
plotting Reference Material results obtained when analysing for gold by fire assay.
This template arose as a response to some of the terrible charts we have seen produced by some LIM's and other software. In
particular, we have seen many hard-to-read charts, based on completely inappropriate process limits, used both by analytical
laboratories, and their customers. We see this poor plotting as less-than-optimal use of our Reference Materials.
We have created this template, based on sound statistical principles, and designed to be of practical use to anyone with a
modicum of knowledge of Microsoft Excel. We hope that it will encourage people to more regularly review the results they are
producing, thus discovering problem results and patterns as they are produced, and then seeking out the causes of the
aberrant results, in order to continually improve the testing.
This blank master template is read-only, and should be used for starting a new chart on any Reference Material. However,
once it is populated with data from a particular Reference Material, it should be saved under a different suitable name.
To save your file (to allow data to be added later) - Select "File" > Save as > (and then for Excel versions prior to 2007 - Tools
> General Options > Deselect "Read-only Recommended" >OK) - then enter a destination path and new filename and click
"Save". This procedure will stop "Read-only" warnings appearing when you re-open the new workbook. Further readings can
then be added to the bottom of the Results Table as they come to hand, and the chart will be updated.
If you are really interested in improving testing, results should be entered directly, as they become available. Reasons
should then be sought for any suspicious points on the chart. The detective work required for genuine improvement will be
much more successful if carried out while the clues are still red hot. Once causes are discovered and understood, the effort
should be made to eliminate them from the testing process.
If you would like to subscribe to (free) e-mailed updates of this template, have any questions
or comments regarding its use, would like help to get started, or experience any problems in
using it, please contact Malcolm Smith at malcolm@MSRML.co.nz
To get started, click the "Start" tab below
Reference Material Plotting Template
Version 2-1.2 May 2010
Select Rocklabs Reference Material that you are plotting ->
(click in box, and then click on down arrow to see a list of RM's, then scroll up
or down to your selection and click on it. The Assigned value and 95%
Confidence limits will then automatically be entered below)
or Enter ID of Reference material you are plotting here ->
(and enter Declared Value and 95% confidence Limits below)
Assigned Value of Standard ->
95% Confidence Limits Standard (+/-) ->
Now click on "Enter Data" tab below
Analysis Table All results
Gross Outliers User Outliers
Number of results 0 #DIV/0! 0 Low data nos. - see note below
Average #DIV/0! #DIV/0! #DIV/0!
Accuracy: (% Difference of Average from Assigned Value) #VALUE! #VALUE! #VALUE!
Precision: Relative Standard deviation (Robust) #DIV/0! #DIV/0! #DIV/0! #DIV/0!
Number of Outlying Results (Outside Process Limits) 0 #DIV/0! #DIV/0!
Perecentage of Outlying Results #DIV/0! #DIV/0!
Note: You have less than 20 results. If you have a historical Relative Standard Deviation, enter % here ->
Comments may not be valid with a low number of results.
Mean Lower Process Limit Upper Process Llimit
Process Performance Chart Production data Standard Gross Outliers 0.000
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140
Explanation of terms
Reference Material Plotting Template Version 2-1.2 May 2010
The Accuracy of your results is measured by the difference between the average of all your laboratory results (after the out-of-co
excluded) and the assigned value, as provided in the Certificate of Analysis that accompanies the Reference Material. The differen
percentage of the assigned value.
Precision is a measure of how variable your laboratory analytical procedure is. This is expressed as a relative standard deviation
deviation of all the in-control results expressed as a percentage of the average of these results. We actually use the median movin
deviation in the template calculation (and refer to it as a 'Robust" RSD) because it tends to reduce the effects of trends than can occ
slightly out of control. The robust RSD is generally lower than the simple, traditional RSD. Standard deviations are by their nature m
averages; nevertheless, RSDs obtained by the same laboratory method on different good quality Reference Materials (no matter wh
trend towards the same value when based on large numbers of results. This is because it is a measure of the laboratory method va
measure of the analyte variation in the Reference Material. A good quality Reference Material should have an inherent analyte vari
laboratory method variation. The Certificate of Analysis provided with Rocklabs Reference Materials has an estimate of the analyte
Homogeneity Assessment section. It is possible to have results that are accurate (have a small difference between average of labo
assigned value), but not precise (have a high RSD). The reverse is also possible: - ie the results may not be accurate but could be
aim should be to produce results that are both accurate and precise.
An Out-of-Control result is any result that is obtained when the normal procedure is not being followed. A Sporadic (or single
result may be grossly different from the average value, and well outside the process limits. Such out-of-control results can have a b
and standard deviation of the remaining laboratory results and should be removed so that the underlying accuracy and precision of
can be measured without being skewed by these extreme values. Typical causes of these sporadic out-of-control results are:
The sample or Reference Material has been wrongly identified.
The weight or volume or dilution factor has been entered incorrectly.
The samples have not been processed in their correct order.
Sporadic Out-of-Control results are normally identified as being results that are well outside +/- 3 standard deviation process limits.
The Percentage of sporadic Out-of-Control results is not necessarily related to either accuracy or precision. Just as it is possible
results that are not precise (and vice versa) it is possible to have an undue number of Out-of-Control results in a laboratory that norm
accurate and precise work. In such cases this is often due to the Reference Material being wrongly identified (either by the laborato
client). We have observed that many commercial and mine laboratories have up to 5%, or even more, of such sporadic out-of-contr
Trends and Shifts might be indicated when results are just outside Process Limits (particularly when a number occur together), o
within the Process Limits, but tending to cluster towards one of the limits. These are best identified by examining the plot of Referen
They are characterized by consecutive results appearing on the same side of the average line, or trending up or down, or a sudden
Possible causes of trends and shifts are:
Deterioration of instrument calibration standards
Instrument drift, changes to furnace temperatures, dispensers, and balances.
Changes in quality of reagents such as flux or cupels.
If the change is large it can be detected after just 2 - 5 results have been plotted. More gradual changes may only be detected after
been plotted. We strongly recommend that every result is plotted as soon as it becomes available and then the plot is examined for
a trend or shift is noticed, the cause should be identified and corrective action taken. Quick action may prevent later grief! The chart
display a message that a possible trend or shift is present if it detects 9 or more consecutive results either above or below the avera
on of terms
ting Template Version 2-1.2 May 2010
ge of all your laboratory results (after the out-of-control results have been
accompanies the Reference Material. The difference is expressed as a
This is expressed as a relative standard deviation (RSD) - ie the standard
these results. We actually use the median moving range standard
t tends to reduce the effects of trends than can occur when the process is
nal RSD. Standard deviations are by their nature more variable than
ent good quality Reference Materials (no matter who produces them) will
ecause it is a measure of the laboratory method variation and not a
ence Material should have an inherent analyte variation that is a third of the
Reference Materials has an estimate of the analyte variability in the
(have a small difference between average of laboratory results and the
: - ie the results may not be accurate but could be precise. The laboratory
dure is not being followed. A Sporadic (or single event) out-of-control
ess limits. Such out-of-control results can have a big effect on the average
d so that the underlying accuracy and precision of the laboratory procedure
of these sporadic out-of-control results are:
ell outside +/- 3 standard deviation process limits.
either accuracy or precision. Just as it is possible to have accurate
er of Out-of-Control results in a laboratory that normally produces both
erial being wrongly identified (either by the laboratory or the submitting
o 5%, or even more, of such sporadic out-of-control results.
mits (particularly when a number occur together), or even when results are
are best identified by examining the plot of Reference Material results.
average line, or trending up or down, or a sudden shift upwards or
More gradual changes may only be detected after 10 - 15 results have
ecomes available and then the plot is examined for trends or shifts. Once
en. Quick action may prevent later grief! The chart will automatically
onsecutive results either above or below the average.