Poll: Accuracy of Test Interpretation

Wrath of the Swarm said:
The statement that an assumption is necessary is a false one.

I just want that to be admitted, and this thread can end peacefully.
Click to expand...

Lets see the maths then. Show me how you can caluculate a unique answer from the information in the intial question.
 
Wrath of the Swarm said:
The statement that an assumption is necessary is a false one.

I just want that to be admitted, and this thread can end peacefully.
Click to expand...
Not the faintest chance. And once we've cleared this up, it's only the start, because it's not what I wanted to talk about anyway.

Now will you please read the post I was writing when you posted this? I can't help feeling that if you'd actually read what I was saying several times before, you'd have got it by now.

You're simply being Humpty Dumpty, making words mean what you've decided they mean, in a field where every primer is prefaced by a list of defined terms in order to avoid misunderstandings like the one you created. By using and sticking to a word which simply isn't defined, because it's a useless concept which would never apply to any but the most flukey real-life tests - then asserting that since you'd used that word, everyone should have known that the test you were using was one of these flukey I've-never-seen-such-a-thing examples.

Rolfe.
 
ceptimus said:
I assert that almost any question, no matter how carefully stated, can be spoilt by pointless quibbling. You almost have to use the language of mathematics to avoid this.
Click to expand...
It would have been nice if this question had been carefully worded. Then we could have moved on to the interesting bit. But it wasn't, in two crucial aspects. And "accuracy" is the simple and boring one.

Wrath, you don't need to use the language of mathematics. You only need to use the already-defined language of the field you are presuming to discuss, as I've pointed out to you too many times already. But by driving headlong with your insistence in your own words and your own meaning, you have created a problem which could have been solved in about four words on page 1.

"Yes, I meant specificity."

There, that wasn't so hard, was it?

Rolfe.
 
That would be a scathing indictment, if only I had used words in ways other than what they were generally accepted to mean.

Test error isn't always affected by the type of response the test gives. In such cases, the test itself can be said to have a specific accuracy.

When the error does change depending on the response, then the test has a specific alpha and beta error rate.

You insisted that such information is necessary to answer the question, when it is not. This behavior on your part demonstrates that you really don't understand the concepts involved, other than in the rote-learning sense.

I'm sure that you can follow a learned algorithm as long as conditions don't fall outside a narrow, prescribed range. But change the nature of the questions presented to you just slightly and... well, we've seen the results.
 
Rolfe said:
Wrath, you don't need to use the language of mathematics. You only need to use the already-defined language of the field you are presuming to discuss, as I've pointed out to you too many times already.
Click to expand...
Translated: "Mathematical terminology frightens and confuses me. I want to stick with terms used in medicine, because I've learned them by heart and can refer to them without having to think and stuff."

[edit] Stupid typos.
 
Wrath of the Swarm said:
You insisted that such information is necessary to answer the question, when it is not. This behavior on your part demonstrates that you really don't understand the concepts involved, other than in the rote-learning sense.
Click to expand...
No, Wrath. Why don't you try again to understand it.

All I said was that I had to assume that when you said "accuracy", you actually meant specificity. I was perfectly prepared to do that, and did so. For the purpose of the question I do not care what the sensitivity is. I do not need to know. I never asked you.

However, you have spent the best part of eight pages asserting that "accuracy" must mean that sensitivity and specificity are the same.

Which is a virtually meaningless concept.

You say you are using the word in its "generally accepted" usage. But the only usage which means anything in this discussion is one defined before you have the discussion. Quite a lot of terms are. A term for "the figure you get when sensitivity and specificity happen to be the same" is not defined anywhere, because it's no practical use.

In fact "accuracy" does have a defined meaning in the context of clinical laboratory analysis, as I posted earlier. It just doesn't happen to be of any relevance to this problem.

Your insistence that we accept that sensitivity and specificity must be equal whenever you choose to use the word "accuracy", rather than simply acknowledging that only a specificity value is needed for the problem and the sensitivity is irrelevant and could be anything, demonstrates that you do not understand the concepts involved, and are trying to handle them with tools developed for rather different purposes, and coming unstuck.

Rolfe.
 
But I didn't mean "specificity", I meant accuracy. It's a more powerful concept with strict standards of application.

Admit it: you are so used to thinking about provided alpha and beta values - named 'sensitivity' and 'specificity', respectively - that you mistakenly assumed your familiarity with them meant they had to be present, and their absense meant a mistake had been made.

But your ego and your need to always be perceived as correct won't let you admit the error, will it?

'Whore' is the wrong word for you. I can respect an honest prostitute. You have no intellectual integrity whatsoever.
 
WOTS: It is a pity that you are so aggressive, because I guess you have some knowledge, and perhaps even some interesting viewpoints. Unfortunately, it is all lost in abrasiveness.

Hans
 
Wrath of the Swarm said:
Translated: "Mathematical terminology frightens and confuses me. I want to stick with terms used in medicine, because I've learned them by heart and can refer to them without having to think and stuff."
Click to expand...

I don't think that the term "accuracy" means what you think it does in mathematics, either.

Semantic quibbles aside : what was your original point going to be?
 
First, my name is not Mr. Fluffer-nutter Fuzzy Wuzzy. It's WRATH OF THE SWARM. If you were expecting kid gloves and gentle talk, you've come to the wrong poster. If you're looking for compassion, you've come to the wrong universe.

Second, if you can't look past the personalities in an argument to see the arguments, you're a fool. You're no different from the believers that get suckered by personalities and pretty speeches and swallow any argument disguised in them.
 
Remember guys, this Wrath fool is the same idiot who said that its OK if people get killed from "alternative" medicine if the gene pool is improved.

Expecting logic to come from such a dolt is an exercise in futility.
 
WOTS: Certainly :rolleyes:, but BESIDES showing off your intrinsic aggressions, do you HAVE A POINT?

Hans
 
Wrath of the Swarm said:
But I didn't mean "specificity", I meant accuracy. It's a more powerful concept with strict standards of application.
Click to expand...
OK, more ad-homs. Enough.

I would like to point out again that I run a diagnostic testing laboratory. That is run. I choose the serological tests we will offer, and decide which is going to perform best according to the demands our clients will make on it. I evaluate tests, and decide which are the most appropriate for our requirements. I have published on the subject, in a refereed journal. Oh, and by the way, not one of these tests has equal sensitivity and specificity.

And every day, I am responsible for "signing-out" scores of serological test results from real patients (and the arithmetic works just the same way in animals as it does in humns). I am also responsible for advising my colleagues in practice on the true reliability of any test result they have received. I make my living doing this. So far, I have not exactly gone bankrupt.

On the contrary, I am considered something of an expert in the field. I have been invited to speak at conferences on the subject. As well as the refereed publications, I have pubished a text-book which included a chapter covering just this issue, and an eminent professor chose that chapter for especial praise in a very flattering book review.

Wrath, we may disagree. But I do not appreciate language like the following.
'Whore' is the wrong word for you. I can respect an honest prostitute. You have no intellectual integrity whatsoever.
Click to expand...
Interestingly, we know nothing of your qualifications, your background in this or other fields, or your reason for being interested (indeed, aerated) about the subject. All we know is that you're so sure you're right you won't even consider any alternative.

Now, to your point above. Please explain in what way a term which is only applicable to the rare and coincidental cases where specificity and sensitivity happen to be the same, is "more powerful" than the standared expressions of sensitivity and specificity. It's certainly a strict standard of application - it's only applicable in rare and unusual cases. But what makes it powerful?

Rolfe.
 
That definition of accuracy obviously does not fit the "accuracy" discussed here. The traditional definition of accuracy (as specified on the referred site) is not dependent on the probability of a given value.

That the "accuracy" of a medical test IS dependent on the probability of a given result is new to me (and seems odd to me) but I bow to the authoritiess here.

Hans
 
Originally posted by Wrath of the Swarm
If alpha != beta, then we can't say the test has an accuracy.

We said the test has an accuracy.

Therefore, alpha == beta.
Click to expand...
If the test is 99% accurate at spotting an affliction with a 1/1000 occurence, the "alpha" value is 0.00001. In other words, 99.99999 specificity. Very unlikely :rolleyes:

Your test is only 9% accurate.
 
Wrath of the Swarm said:

http://mathworld.wolfram.com/Accuracy.html

Nice try. Well, it wasn't actually a very good try at all, but you get points for effort. Well, since it was a throwaway attempt at a stab in the dark, no you don't.
Click to expand...
Not a relevant definition to the subject under discussion. In fact it's a general form of the definition I posted earlier which is used in the biochemistry field. Note the coupling with the related concept, precision.

However, the only terms you can use to discuss the very precise field you are involved in now are those which can be derived from the various permutations of the numbers of true-positives, false-positives, true-negatives and false-negatives in your test population.

If you can't define your term that way, it is not a useful term.

I'm not throwaway stabbing in the dark, I'm simply requiring you to do what I require any student to do before discussing the topic. Understand the definitions of the terms used, and be able to define mathematically with reference to these four figures what any other term they want to use actually means.

Rolfe.
 
MRC_Hans said:
That the "accuracy" of a medical test IS dependent on the probability of a given result is new to me (and seems odd to me) but I bow to the authoritiess here.
Click to expand...
No, Hans, it isn't. "Accuracy" as Wrath is trying to define it, is simply not a useful concept in this field. Which is why the word isn't used.

Specificity and sensitivity are not dependent on the probability of a given result. However, it has been found useful also to consider the positive and negative predictive values, which are.

It's explained back where I quoted BillyJoe's explanation, because he was kind enough to deminstrate how all the terms we use are derived from four basic figures derived in the initial evaluation of the test. Wrath's "accuracy" cannot be so defined, therefore it is not a useful concept.

Rolfe.
 
But if you always assume that what an "authority" says is true, you'll never find evidence that what they say isn't actually correct.

The degree to which a test's results match the actual values of the test subjects is the test's accuracy. The standard definition given in the link is precisely what we're talking about - we're discussing binomial distributions, not continous ones. The degree to which something is free from error is its accuracy. If we can say that the test's results will match reality 99% of the time, regardless of the nature of the result itself, we can say that the test is 99% accurate. If the specific result matters, we can't say that - we need to specify alpha and beta values.

The accuracy of the test is one thing; the accuracy of the conclusion drawn from that test's result is another. That's the whole point of the question - it's counterintuitive.
 
Originally posted by ceptimus
No exarch. alpha or beta can't be zero. Here's a simple way to see why.

Say everyone in the population has the disease. The test is 99% accurate remember.

Now say no one in the population has the disease. The test is 99% accurate remember.

As regards the unambiguous question, post a copy of it here, and we'll see if we can find a quibble.
Click to expand...
You're right. if the accuracy is the occurence of false positives, it can't be.
Ceptimus, this whol thing hinges on WotS's misunderstanding between "accuracy" and "specificity" (occurence of false positives).

WotS's test is in essence only 9% "accurate" at positively identifying someone as having the affliction, not 99%.
It is 99% specific though.
 
You must log in or register to reply here.

Back
Top Bottom