I understand that, but the main problem I have is that the data therein is subject to the same concerns I had about grabbing numbers off of data sheets, much less safety sheets. We have singular data points, with no error analysis (that we can review). This isn't the same thing as grabbing numbers from someone else's raw data and working with their error analysis or doing your own. We're dealing with "typical" values, as the wording usually is, without either raw data or data dealt with by our own error analysis or one we can review. I do understand that such information isn't going to be easily forthcoming, but that doesn't change the scientific method.
Look at ordinary experimental procedure. How many times was KV40 or KV100 run by the oil company on a particular batch? I certainly don't know that. Yes, I have confidence that the value they give is typical and close enough for most purposes. Is it close enough when you're trying to establish a scientific relationship? That's where I have a problem.
Also, not to say that data sheets are useless, but there are enough errors here and there, some very glaring, that we have to watch. If there are glaring errors, how many are almost undetectable?
The predictability idea is thus, as you know. You should be able to be handed several bottles of oil, and conduct the tests - several times each - to fill in the data. Then, the calculation and predictions should hold true.
I'm the first guy to like theory and dislike the lab. That doesn't mean I can discount the value - the necessity - of experimentation. There's a reason why people once thought that heavier objects fell more quickly than lighter objects - the careful experimentation wasn't done.
Look at ordinary experimental procedure. How many times was KV40 or KV100 run by the oil company on a particular batch? I certainly don't know that. Yes, I have confidence that the value they give is typical and close enough for most purposes. Is it close enough when you're trying to establish a scientific relationship? That's where I have a problem.
Also, not to say that data sheets are useless, but there are enough errors here and there, some very glaring, that we have to watch. If there are glaring errors, how many are almost undetectable?
The predictability idea is thus, as you know. You should be able to be handed several bottles of oil, and conduct the tests - several times each - to fill in the data. Then, the calculation and predictions should hold true.
I'm the first guy to like theory and dislike the lab. That doesn't mean I can discount the value - the necessity - of experimentation. There's a reason why people once thought that heavier objects fell more quickly than lighter objects - the careful experimentation wasn't done.