Based on the majority of Jano testing, the gcms appears to be fairly accurate.
Very important typo. I meant to say
fairly inaccurate. Thank you
@readalot for bringing this to my attention.
What do you mean by this statement? Could you elaborate please?
Long story short, I decided to do some basic statistics. Note the uncofindence in my statement — these kinds of statements are closer to "more likely than not" rather than "95% confidence". This is mainly due to the small sample size (~10) more than anything. Again, the conclusion is supposed to be that GCMS area is
not accurate rather than is.
Longer explanation:
We have two methods that are essentially measuring purity, some more accurate than others. If we make three basic assumptions, we can do some comparative statistics like lims of agreement. We do this by making a table of GCMS vs HPLC data (of the same samples) and then comparing.
Assumptions:
First: All samples are thoroughly mixed and consistent throughout, especially when one individual sends two samples for testing. This is less likely with bodybuilders rather than actual researchers, but still expected.
Second: Jano's hplc is indeed very accurate, other than confusing very similar compounds like d4,6.
Third: There are no significant quantities of involatile substances. This is the least certain, but is still likely due to how organic synthesis works practically.
If so, we can create a spreadsheet with 2 columns. For the HPLC column, we just write the purity. For the gcms, we add any compounds' areas that will almost certainly coelute on hplc (in general this is only testosterone and d4,6 - see below for more info). Then we just compute the residuals, do some altman analysis, etc. While the statistical power is lacking with only n=~10, it can still tell us a lot. Honestly, if anyone here has any time, please make a spreadsheet of any janoshik reports that you have of the
same sample, with gcms
and hplc. If we can get enough data, we could even do cluster analysis and find patterns with Jano's testing, possibly even with vendors.
How did you come to this conclusion?
This is a bit less scientific, and is mostly a combination of: how likely two molecules are to coelute based on structure (in the case of d4,6, I would say the damning aspect is that the double bond is in very close proximity to the d4 bond, unlike modifications far away from it like boldenone or nandrolone) and published data, how the hplc raw data looks (especially the chromatograms, I actually think that I saw a shoulder peak on one), how it compares to the gcms, etc. Its a bit hard to explain, but at minimum I haven't seen anything nearly as flagrant as the d4,6 and testosterone.