Does the use of error bars pertain to this experiment?

[hydronicengr]

I'm currently performing a real bench experiment that would essentially validate the results of those generated in two different computer models.

Essentially, there is one (real) test setup, one computer model, and another (but different) computer model.

I have ONE result from each of the tests, so in total I have three numbers. The real test setup and one of the computer models are similar to each other at 2.70 and 2.71, respectively. The other computer model is 2.80.

I want to see if the last computer model result of 2.8 is "acceptable" or falls "within range" of the other results.

I was told to use error bars, but I'm not sure if this would be the best approach.

 

[Stephen Tashi]

I have ONE result from each of the tests, so in total I have three numbers.

I want to see if the last computer model result of 2.8 is "acceptable" or falls "within range" of the other results.

I was told to use error bars, but I'm not sure if this would be the best approach.

If you have one result from each test and this result is not the average of many other replications of the test, then I can't think of any reasonable interpretation of your question which has a convincing mathematical answer.

Perhaps you have a sociological problem, not a mathematical problem. Are you preparing a report or briefing or article that will be evaluated by other people. If so, you should look at other reports or briefings that they liked and copy whatever methods were used.

Your request to determine whether 2.8 is "acceptable" doesn't define a specific mathematical problem. The best I can do to mind read what you want is to day that you want to assume that the number 2.70 and 2.71 are drawn independently from the same Normal distribution and you want to compute an estimate of the mean and standard deviation of this distribution. Then you want to draw a bar whose center is the estimated mean and whose width is a certain number of the estimated standard deviations. You can certainly do that. It isn't a mathematical proof of anything.

If two of the numbers come from deterministic models, it would require some verbal contortions to explain how they come to be viewed as "random" samples. I suppose you could claim that the decisions made in creating the simulations involve some random choices.