Chromatography Forum

In a GC/FID analysis using straight-chain aliphatic hydrocarbons as standards, I were to determine whether the calculated ECN of my chosen compound (1-Decanol) is statistically different from a straight-chain hydrocarbon with the same number of carbon atoms provided that each carbon in the straight chain hydrocarbon has ECN of 1.0?

The aliphatic hydrocarbons used in method development were heptane, octane, nonane, decane, undecane, dodecane.

In analysis of 1-Decanol, I used undecane and dodecane.

I got 9.8 for ECN of 1-Decanol. Does that mean "Yes, statistically different since it is not 10.0?" I will also be calculating 95% confidence level.

Hi

Decane is C10H22 EC= 12*10/(12*10+22*1)=0.83
Decanol is C10H22O EC= 12*12/(12*12+22*1+16*1)= 0.77

0.77/0.83= 0.93
Its not statistic problem- its realy chemistry .

Decanol do not alkane [hydrocarbones] gomolog , its "highboiling one atomic spirit ", forgot how to call it in English literature
Google sayd " high-molecular monohydric alcohol"

Hi Sandra

I was wondering how you were getting on with this.

How did you calculate the ECN for decanol ?

The question of statistical significance turns on the variability in the results - and not only multiple injections of the same solution (presuming here that you had all your standards and the decanol dissolved together and injected together), but also replicates of the the various weighings and dilutings that went on to prepare the solutions.

Peter

Hi Sandra

I was wondering how you were getting on with this.

How did you calculate the ECN for decanol ?

We are doing another lab experiment but the report for the GC one is due today.

I used dodecane as the internal standard to normalize peak areas. Then, for the standard, to determine whether FID responds to carbon consistently or not, I first take the ratio of normalized peak area to ppm of carbon in the six alkanes, and then the result of these ratios were checked to be within 15% of each other. It's 5.8% between the highest and the lowest ratio (of normalized peak area/ppm of C in alkane) result.

The question of statistical significance turns on the variability in the results - and not only multiple injections of the same solution (presuming here that you had all your standards and the decanol dissolved together and injected together), but also replicates of the the various weighings and dilutings that went on to prepare the solutions.
Peter

Yes, three replicate analyses were done for the standards as well as for the sample containing decanol. With the latter analyses, I were to determine the effective carbon number for 1-decanol (X1). So I set up like this:
X1/Y1 = X2/Y2 where

X1 = ECN for 1-decanol; X2 = ECN for decane
Y1 = Ratio of normalized peak area/ ppm of carbon in 1-decanol
Y2 = Ratio of normalized peak area/ ppm of carbon in decane

I got X1 to be 9.88. 9.57, and 9.98. Average is 9.81. Now I need to make a conclusion statement and I am confused about whether the answer is yes or no.
Since there are ten carbon in 1-decanol, according to the definition given by the professor in the handout (as stated in my initialpost), ECN for 1-decanol would be 10, wouldn't it? In fact, I do not understand the way EC was calculated for 1-decanol in the response by DSP007 (Decanol is C10H22O EC= 12*12/(12*12+22*1+16*1)= 0.77 versus
Decane is C10H22 EC= 12*10/(12*10+22*1)=0.83).

Why 12*12 for the carbon in decanol? Please help.

Hi Sandra

You are still using ppm instead of proper units, so I cannot check you calculations.

Taking your three values for ECN as given, calculate their standard deviation, and then look at whether the mean is significantly different from 1 at the level of significance that you want.

I don't have any clue at all what it is that DSP007 is trying to calculate.

Peter

Sorry, had a great disservice
Now understood what was meant
The essence of error is that in Russian (for FID detector) - abbreviation of the name corresponds to (among other values) and response of carbon "in a number of homologues

Now, seeing the numbers 9.88, 9.57, 9.99 and realized that "do not understand the question"
It was intended interpretation of this result

I'll know that the decline here is completely different meaning.


In fact the question
1) with N = 3 9.57 in this sample is mathematically impossible to identify the clanger (mistify) . (9.57-9.88) / (9.99-9.57) = -0.73 Q<0.94scale deviations (for N = 95% )
But the 9.57-suspicious value.
2) RSD = 0.21, confidence interval, P (r = 0.95, n = 3) 0.25, ie 9.8 in this spread fell into the top ten.
3) In terms of formal mathematics 9.5 is rounded to 10, but not up to 10.0

Practical
Makes repeat calibration on a large number of samples.
Small errors sometimes lead to serious consequences.

Sorry, had a great disservice
Now understood what was meant
The essence of error is that in Russian (for FID detector) - abbreviation of the name corresponds to (among other values) and response of carbon "in a number of homologues

Now, seeing the numbers 9.88, 9.57, 9.99 and realized that "do not understand the question"
It was intended interpretation of this result

I'll know that the decline here is completely different meaning.


In fact the question
1) with N = 3, 9.57 in this sample is mathematically impossible to identify the clanger (mistify) . (9.57-9.88) / (9.99-9.57) = -0.73

Got it so far.

Q<0.94scale deviations (for N = 95% )

Where did Q<0.94 come from?

But the 9.57-suspicious value.

Yeah

2) RSD = 0.21, confidence interval, P (r = 0.95, n = 3) 0.25, ie 9.8 in this spread fell into the top ten.

In the above, 0.21 is Std Deviation, right?

3) In terms of formal mathematics 9.5 is rounded to 10, but not up to 10.0

OK

Practical
Makes repeat calibration on a large number of samples.
Small errors sometimes lead to serious consequences.

OK. Problem was that we had limited amount of time.

Hi Sandra

You are still using ppm instead of proper units, so I cannot check you calculations.

Sorry. People looked at me funny in class if I don't.

Taking your three values for ECN as given, calculate their standard deviation, and then look at whether the mean is significantly different from 1 at the level of significance that you want.

OK. I can find Std Dev and then find c=Confidence Level that way.

The prof. also said that the other way is to calculated the uncertainty. Since I have 3 sets of peak areas (of the same sample prep) for undecane, dodecane, and 1-decanol, if I use either undecane or dodecane peak areas as y values - x values would be carbon number of that alkane, do I take the average peak area of 1-decanol as measured y and do Linear Regression Analysis in order to find derived x value for 1-decanol? Only then will I get m (slope), Sm (Std. Dev for m), Sy (Std. Dev for y), etc. in order to calculate Sx (uncertainty in x) using a formula for Sx.

I have two formulas to calculate Sx. I'll ask the prof. today. I am turning in my report late. Points will be deducted but I am not a degree seeking student and so I am not worried about my grade.

I don't have any clue at all what it is that DSP007 is trying to calculate.

Peter

Hi Sandra

You are still using ppm instead of proper units, so I cannot check you calculations.

Sorry. People looked at me funny in class if I don't.

Taking your three values for ECN as given, calculate their standard deviation, and then look at whether the mean is significantly different from 1 at the level of significance that you want.

OK. I can find Std Dev and then find Confidence Level that way.

Well, I checked with the prof and now I am clear what he wants on the Statistics.

Hi Sandra

If you are going to calculate uncertainty then you need an uncertainty budget for the whole process - starting with certified values for the purity of the reference materials that you used. I seriously doubt that this is what your Prof has in mind.

You should refer your class mates, and your Prof to the IUPAC Green Book - it is very clear that ppm and similar dimensionless ratios should never be used.

Peter