how much air in the tank?

[Crush]

Perhaps this is just an issue of symantics. But standard error is a statistical measure measuring the variance in the data from repeated sampling or measurement. That you did a calculation, means there is only one data point. If you had done simulations or experimentations and had multiple data points, then you could provide standard error. With your method, I think your best bet is just to stick to significant figures.

It is not "symantics." You are arguing with me without understanding what I am writing. You clearly don't know what I meant by a Taylor series approximation of the errors or you wouldn't be typing such nonsense. The standard deviation of a single data point from an equation can be estimated via a Taylor series approximation.

We are not at the high temperatures where the breakdown of the ideal gas law is sufficient to warrant a new model. Its been over a decade, but if memory serves, for the pressures and temperatures involved, ideal gas law is quite accurate and the model engineers use.

I give up. You have clearly completed more courses in physical chemistry and chemical physics than I have. :sarcasm:

I'll be staying off this thread from now on. If you want to continue to teach me statistics and physchem from your unique perspective, PM me.

 

[watboy]

No need to get upset. I've fully qualified my lack of knowledge regarding gas modeling, so more than willing to concede that point. But statistics is something I do know a thing or two about. And yes, I clearly don't understand what you are writing. You said you observed values, when you only ran calculations. You stated a +/- 1% error, without stating a confidence interval. For learnings sake:

1.0 + 1.0 = 2.0

Whats the standard deviation on my answer? I'm only partially being difficult on this, if there's a way to calculate the standard deviation of my answer with just the above, i'd love to know how to do it. I would think that knowing that my parameters only have 2 significant digits, does not reveal anything about the underlying distribution. And without that, whatever error function I use, or whatever infinite series i need to approximate it, I don't see how one would get a standard error. Again, I could be wrong, happy to hear your explanation.

It is not "symantics." You are arguing with me without understanding what I am writing. You clearly don't know what I meant by a Taylor series approximation of the errors or you wouldn't be typing such nonsense. The standard deviation of a single data point from an equation can be estimated via a Taylor series approximation.



I give up. You have clearly completed more courses in physical chemistry and chemical physics than I have. :sarcasm:

I'll be staying off this thread from now on. If you want to continue to teach me statistics and physchem from your unique perspective, PM me.

 

[Crush]

OK, one last post.

You need to learn more about statistics before typing more rubbish. See Propagation of uncertainty - Wikipedia, the free encyclopedia .

In the case of 1.0 + 1.0 = 2.0 the implicit uncertainty in the quantities being added is 0.1 (as it has not otherwise been stated) so the uncertainty in 2.0 is Sqrt[0.1^2 + 0.1^2] = Sqrt[0.02] = 0.1414... ~= 0.1 where "~=" means "approximately equal to."

 

[spoolin01]

mpetryk you have brought an interesting discussion, but watboy's comments on the statistics were clearly valid enough not to be termed rubbish so harshly. His approach was a conventional empirical take on the concept of error - the relation of a population distribution to precision - and since you went beyond sig fig (which is at best a loose means of expressing error and you acknowledged you were just assuming the significance of the coefficients) to invoke standard error, which has an accepted and commonplace stochastic meaning, his challenge was understandable. As an aside, I don't think it's common for sig fig treatment to be used to capture error, in fact it's almost never used in the areas I've worked in, except loosely to control runaway precision in the output. I think it's far more likely those coefficients are simply truncated where they are for arbitrary reasons, than that they embody the error of the model, simply because 1 part in 300 or 1000 is more than enough precision for an expression that merely approximates something. If you were trying to land a man on the moon, you might need better than that.

So again this begs the question: how appropriate is the VDW model? As a reference, it doubled the deviation from ideal that Harlow presented in his graph for pressures below the cross-over point. Assuming for argument that the Harlow graph for air is accurate, the ideal and VDW models were about equally good at predicting volume from pressure up to 3000 psi. Any idea why the discrepancy?