Glossary

These are the terms used in QC Analytical Results and this manual.

These are the terms used in QC Analytical Results and this manual. How this is calculated, at the top left of each chart, gives the formulas.

Account
Your Geoscience Analytics Toolkit account. You sign in with it to use your own data in QC Analytical Results, and the demo needs none. It saves the Overview rules you set, so they are there on any computer. “Signed in as” in the header opens your account page.
Accuracy
How close results are to the true value. QC Analytical Results checks it with certified reference materials (CRMs).
Action limit
The line beyond which a result is out of control. On Shewhart charts it is 3 standard deviations from the expected value. For duplicates it is twice the check stage’s precision limit.
Analysis suite
The laboratory’s method or package of analyses that a result was reported under. The app keeps results from different suites apart.
Analyte
The element, method and units of a result, as the laboratory names it, for example Cu_MEICP41s_ppm.
Batch check
The Lab files tab’s verdict on one laboratory certificate before it goes into your database. The verdict is pass, query or fail, and each finding says what is odd, why it matters and what to check. The check compares the sample count with the number declared. It also looks at the lab’s duplicates, repeats, blanks and standards, values that aren’t results, codes and prefixes still to confirm, and detection limits. With your QC sheet, it checks your own CRMs, blanks and duplicates too.
Bias
A steady difference between a CRM’s results and its certified value, where the laboratory reads high or low. It is usually given in standard deviations or as a percentage.
Below detection
A result smaller than the laboratory can measure. It is written as <0.01, as minus the LLD such as -0.01, as exactly half the LLD, or as any other value under the LLD. The app leaves it out, so it is never charted, never in a statistic, and not mentioned on the chart. A duplicate pair with one is left out as a pair. Blanks are the exception. A blank below detection is a clean blank, kept at half its limit and drawn as an outlined bar. A negative made only of 9s, such as -99, -999 or -9999, is not below detection. It is a no-data code, read as missing.
Blank
Material with almost none of the analyte, sent with the samples. A blank that reads high points to contamination.
Certificate
The laboratory’s report of a batch of results, as a file. It holds the certificate number and its status, and header rows with the client, number of samples, dates received and completed, and PO number. Then come the method, element, unit and detection limit of each column, and one row per sample. QC Analytical Results reads SIF certificates, comma-separated or fixed width.
Certified value
A CRM’s accepted value, with its certified standard deviation, from the CRM’s certificate.
Change point
A result after which a CRM’s results shift. Change-point analysis finds it and gives a confidence. At 95% or more it is a change, and at 90 to 95% a possible one. Tick Show change points on the Shewhart or CUSUM chart to see them.
Check stage
The stage a duplicate was taken at, such as field, crushing or pulverising. The Overview and RD Rolling Median check each stage against its own limit. By default the limits are 30% for field, 20% for coarse and 10% for pulp duplicates.
Client
For a consultancy, a company whose QA/QC it looks after. Each client has its own projects, CRM library and rules for what fails a batch. Every client shares the labs’ codes and prefixes, unless one needs its own.
Contamination
The analyte getting into a sample from equipment or from the sample before it. QC Analytical Results checks it with blanks.
Control lines
The lines on a chart that a result is checked against, such as the warning and action limits.
CRM (certified reference material)
A material of known, certified grade sent with the samples to check accuracy. Company standards are bought or made by the company. Laboratory standards are the laboratory’s own.
CRM library
Your standards and their certified values, SDs and acceptable ranges. It is saved with your account or a client, and you add to it from your database’s standards table or the template. The Lab files tab checks the lab’s standards and your CRMs against it.
CUSUM
Cumulative sum chart. It adds up each result’s difference from the target, so a small, steady bias builds up and shows before any single result fails.
Despatch
A batch of samples sent to the laboratory together, under one despatch number.
Duplicate
A second analysis of the same sample, taken at a check stage. The two results are a pair, Sample 1 and Sample 2, and neither is treated as the original. Their difference measures precision, and part of it can come from the geology.
EWMA
Exponentially weighted moving average chart. Each point averages the latest result with the ones before it, with recent ones counting most, to show gradual drift.
Fold
A blue line with a ▸ in front, such as Show 5 more, that hides more rows. Click it to open them, and the ▸ turns to ▾. Click it again to close them.
Holding area
Where certificates kept in a project wait, checked, until you accept them or return them to the lab. Accepted results then open with the project.
Lab Confidence
The Overview’s overall rating of a laboratory: High, Moderate, Low or Insufficient data.
Lab duplicate
A second analysis the laboratory makes of a sample. It is reported as an extra row named with a prefix and the sample number, such as Ch:307522 or Dup:EXPPT0001. It is checked against the limit for the stage you give its prefix.
LLD (lower limit of detection)
The lowest value the laboratory’s method can report. Results below it are left out, except for blanks. See Below detection.
MAD (median absolute deviation)
A measure of spread that a few extreme results don’t distort. QC Analytical Results can draw control lines from it instead of the standard deviation.
Nelson rules
Eight tests for patterns that show a process is out of control, even when every result is within its limits. One example is a run of results on one side of the mean.
Precision
How well repeat analyses of the same sample agree. QC Analytical Results checks it with duplicates.
Project
Your QC data saved in your Toolkit account, by project, and by prospect if your files have a PROSPECT column. After an upload, Save adds the file to a project as an upload. New lab jobs are added, and a lab job the project already has is replaced whole by the file’s results, after the app asks. The project comes from a PROJECT column in the file, or you choose it. The Projects data source opens a whole project or one prospect, and signing in opens the project you changed last. Only you can see your projects. Each stays until you delete it in Manage projects, under the Projects button, or close your account. An account holds up to 20 projects or 1 GB for now.
QC sheet
A list of which of your samples are your own QC samples. Its columns are Sample, Kind, Standard and Original, where Kind is CRM, blank, field, coarse or pulp duplicate, or umpire. A database’s check-sample export works as it is. With it, the Lab files tab checks your own QC samples too.
RD (relative difference)
The difference between a duplicate pair as a percentage of their size. QC Analytical Results offers several forms: ARD%, CV%, HARD%, MPRD% and RDHYSLOP.
RD Rolling Median
The median relative difference of the duplicates in each despatch and the ones before it, over a rolling window. It shows whether precision is changing.
Re-issue
A certificate the laboratory sends again, corrected, under the same number. If you keep it in a project, it replaces the old one after you agree. The old results are kept but no longer used.
Repeat
On a certificate, an extra column of the same element, method and unit, such as Au1 and Au2, or Au and Au-Rp1000. It is the same pulp analysed again, checked against the instrument stage’s limit.
Result code
A result that isn’t a number, such as X, IS or LNR. A code can mean below detection, above the upper limit, not assayed, insufficient sample, not received or missing. The meaning differs between labs and sites, so the app asks once and remembers.
Row prefix
The laboratory’s name for its own check rows, such as Ch and Dup for duplicates, St and Std for standards, and Blank. The app asks once what each prefix is, and the stage for a duplicate prefix.
Shewhart chart
Each CRM result plotted in order against its expected value, with warning and action limits.
SIF
The laboratories’ standard interchange file for results. It is a certificate as comma-separated values or fixed-width columns.
Standard deviation (SD)
The usual measure of spread around the mean.
Thompson-Howarth
A precision chart that plots how the spread of duplicate pairs changes with grade.
Threshold factor
A factor that multiplies the LLD to give a threshold. Duplicate pairs below the threshold are greyed out and left out of the statistics.
Ticket rule
A rule that says which sample numbers are your QC samples, for ticket books and bar codes. For example, numbers ending 00 are CRMs, or the ticket after 25 is a duplicate of 25. The app proposes one from the QC samples you have already marked.
Turnaround
The working days, Monday to Friday, that a despatch took from its send date to its last return date. The Overview’s Turnaround card compares each laboratory with its own history. It doesn’t change Lab Confidence.
ULD (upper detection limit)
The highest value the laboratory’s method can report.
Upload (to a project)
One file saved to a project. Each upload is kept, so taking one out in Manage projects rebuilds the project from the others, and a lab job it replaced comes back. The Overview’s Since last upload shows what the latest upload changed.
Warning limit
The line 2 standard deviations from the expected value on Shewhart charts. About one result in twenty falls outside it by chance.
Z-Score
How many standard deviations a result is from the expected value: (result − expected) ÷ SD.