{"id":50,"date":"2026-08-23T15:37:08","date_gmt":"2026-08-23T15:37:08","guid":{"rendered":"https:\/\/qoy.vxf.mybluehost.me\/website_11aa7f22\/user-guide\/"},"modified":"2026-09-17T06:03:59","modified_gmt":"2026-09-17T06:03:59","slug":"user-guide","status":"publish","type":"page","link":"https:\/\/aquaphotomics.com\/software\/aquagrammar\/user-guide\/","title":{"rendered":"How AquaGrammar works"},"content":{"rendered":"\r\n

Short answers, in the order the workflow applies them. If you arrived by clicking an i<\/code> button in AquaGrammar, you have landed on the section that answers it.<\/p>\r\n\r\n\r\n<\/a>\r\n\r\n\r\n

SNV, and the three ranges<\/h2>\r\n\r\n\r\n

SNV (Standard Normal Variate)<\/strong> normalizes one scan against itself: it subtracts that scan's own mean and divides by its own standard deviation. The effect is to remove differences in overall level between scans, so that two spectra can be compared by *shape* rather than by how bright they were.<\/p>\r\n\r\n\r\n

Preview applies SNV to every selected scan automatically. The one choice you make is which wavelength range sets the mean and standard deviation. In every case the whole spectrum is returned \u2014 the range chooses what sets the centre and scale, not which wavelengths survive.<\/p>\r\n\r\n\r\n

Why water has these regions at all<\/h3>\r\n\r\n\r\n

Water's near-infrared spectrum has four main maxima, near 970, 1190, 1450 and 1940 nm. Working outwards from the OH stretching vibration, they correspond to its second overtone, a combination of the first stretching overtone with the OH bend, the first stretching overtone itself, and a stretch-plus-bend combination.muncan<\/a><\/sup> Two of those fall inside the range a MicroNIR records, which is why AquaGrammar offers the choices it does.<\/p>\r\n\r\n\r\n

The three ranges<\/h3>\r\n\r\n\r\n

1st overtone, 1300\u20131600 nm \u2014 the default, and the published convention.<\/strong> The twelve Classic-12 coordinates were found experimentally in the water first overtone and all sit between 1342 and 1512 nm.muncan<\/a><\/sup> This region is used because the water bands resolve clearly there and their assignments are well documented.muncan<\/a><\/sup> The control is labelled with the conventional band. The window itself takes in every measured point between 1298 and 1601 nm, the closest a MicroNIR grid comes to those bounds, and \u03bc and \u03c3 come from exactly the columns inside it, so the precise window is part of the method rather than a display detail. The established aquagram method applies scatter correction across this same region before the per-wavelength standardization,molecules2020<\/a><\/sup> so choosing it reproduces standard practice. Start here.<\/p>\r\n\r\n\r\n

Combination band, 1100\u20131300 nm.<\/strong> The band centred near 1190 nm in this window combines the first overtone of the O-H stretch with the fundamental O-H bend (2\u03bd\u2081,\u2083 + \u03bd\u2082).muncan<\/a><\/sup> It is a weaker<\/strong> absorber than the first overtone, and that weakness is the point: it is the region aquaphotomics turns to for deep-tissue penetration and short-wavelength NIR monitoring, where a standard 1 mm or 2 mm path length at the stronger bands would saturate the sensor. Reach for it when your sample is thick or strongly absorbing enough that the 1450 nm region overwhelms the detector.<\/p>\r\n\r\n\r\n

As a normalization reference it has a second advantage: it lies *outside* the coordinates being plotted. Work on preprocessing for aquaphotomics recommends estimating scatter from regions where the chemical bands of interest are absent, and names 1000\u20131300 nm as one such window.roger<\/a><\/sup><\/p>\r\n\r\n\r\n

(The app's 1100\u20131300 nm window brackets the band generously; the absorption itself is concentrated around 1150\u20131250 nm.)<\/p>\r\n\r\n\r\n

Full range \u2014 every wavelength in the file.<\/strong> Useful for exploration and for comparing against other tools, but be aware of the trade-off: full-spectrum SNV spreads variation belonging to the compound of interest across the whole spectrum, which the same work describes as detrimental when you are trying to recover specific spectral components.roger<\/a><\/sup> It also includes the noisy extremes of the detector.<\/p>\r\n\r\n\r\n

Choosing one for an unfamiliar system<\/h3>\r\n\r\n\r\n

One criterion is concrete: if your sample is thick or strongly absorbing enough that the first-overtone region saturates the detector, work in the combination band.<\/strong> That is a property of your measurement, not a matter of taste.<\/p>\r\n\r\n\r\n

Beyond that there is no rule, and no published decision procedure. When you are exploring a target system you do not know well, the standing advice in the field is to look at each region rather than assume one: some regions suit some aqueous systems better than others, so it is worth examining them.muncan<\/a><\/sup> Practically, run the same selection through more than one range and compare the resulting Aquagrams \u2014 if your groups separate under one range and not another, that itself is worth understanding before you draw conclusions.<\/p>\r\n\r\n\r\n

Do not treat any of the three as "the correct answer". Record which one you used; it is part of the result, and Preview writes it into the provenance line and every processed row's formula for exactly that reason.<\/p>\r\n\r\n\r\n<\/a>\r\n\r\n\r\n

Standardization<\/h2>\r\n\r\n\r\n

After SNV, each wavelength is standardized across every scan in the file<\/strong>:<\/p>\r\n\r\n\r\n

(value \u2212 mean) \/ SD\r\n<\/code><\/pre>\r\n\r\n\r\n
Where SNV works along one scan, standardization works down one wavelength. It puts every coordinate on a comparable footing, so that a coordinate with naturally large values does not dominate the chart's shape.<\/p>\r\n\r\n\r\n
A wavelength that is identical in every scan has no spread; its standardized value is zero rather than an error.<\/p>\r\n\r\n\r\n
Repeat averaging<\/h2>\r\n\r\n\r\n
Finally, the repeats within each group are averaged into a single row. A group scanned fifteen times becomes one line, not fifteen. This is why selecting the right groups in step 2 matters: averaging happens *within* a group, never across groups.<\/p>\r\n\r\n\r\n<\/a>\r\n\r\n\r\n
Operations: Average, Add, Subtract<\/h2>\r\n\r\n\r\n
These act on prepared rows, after everything above has run.<\/p>\r\n\r\n\r\n
Average<\/strong> \u2014 the mean of two or more ticked rows. One representative line from several.<\/p>\r\n\r\n\r\n
Add<\/strong> \u2014 the sum of two or more ticked rows.<\/p>\r\n\r\n\r\n
Subtract<\/strong> \u2014 tick any number of rows to adjust, and choose one baseline<\/strong> row separately. Each adjusted row produces its own result: adjusted minus baseline. Ticking three rows against one baseline creates three results in a single step. A row subtracted from itself produces zero, which is a useful check that you have the operands you think you have.<\/p>\r\n\r\n\r\n
Two things worth knowing:<\/p>\r\n\r\n\r\n