It doesn't become a test because you call it so. A test is something that has the power of testing.
By the power of testing... I have the power!
Again, "test statistic" is just a description of any statistic that you use in a test. There's no other property required. The mean, a coefficient, a t-value, an F-value -- use any of these in a test and they become a test statistic. The fact I've had to explain this simple thing about 25,000 times is a real testament to the molasses-like slowness of these people.
It doesn't become a test because you call it so. A test is something that has the power of testing.This is so wrong I can't tell if it's trolling or not. But if you're serious, no. Just amazingly wrong. Please just read Casella & Berger or anything and send your apologies.
So if I compare my t-statistic to the current dew point in Manhattan, is that a test because I call it a test and I'm comparing two numbers?
You're testing a hypothesis, dumbass, not the statistic itself. And that's only the beginning of what's wrong with your post. I can't believe ignorant people like you exist, but in REP, well...
As explained at length in prior threads, a test statistic is any statistic that you test.
God are you dumb. If you compare the statistic to something, that's the hypothesis test. Are you seriously so stupid that you think it matters that you formulate a hypothesis beforehand? Like you think part of the definition of test statistic involves a sequence of events? Are you actually that dumb?
You need a hypothesis test to have a test statistic. The hypothesis test is formulated in advance.
Hahahahaha. Really? So if I look at a t-value, then formulate a hypothesis test, it ceases to be a test statistic? It's like a quantum measurement thing in your mind?The hypothesis test has been worked out in advance by the authors of the tables and your statistics program. The test statistic is calculated by assuming a distribution under the null hypothesis.
Dude, why are you posting when you clearly have no idea what you're talking about? You're confusing a test statistic with critical values for a specific kind of test. Also, authors of the tables? What?
Again, "test statistic" is just a description of any statistic that you use in a test. There's no other property required. The mean, a coefficient, a t-value, an F-value -- use any of these in a test and they become a test statistic. The fact I've had to explain this simple thing about 25,000 times is a real testament to the molasses-like slowness of these people.
It doesn't become a test because you call it so. A test is something that has the power of testing.
This is so wrong I can't tell if it's trolling or not. But if you're serious, no. Just amazingly wrong. Please just read Casella & Berger or anything and send your apologies.So if I compare my t-statistic to the current dew point in Manhattan, is that a test because I call it a test and I'm comparing two numbers?
If 1 is bigger than 2 then you cal it like it is!
This is so wrong I can't tell if it's trolling or not. But if you're serious, no. Just amazingly wrong. Please just read Casella & Berger or anything and send your apologies.
Ugh. If this is in reference to the Casella and Berger piece I’m thinking of, while it wasn’t outright retracted it was so heavily amended later that it really shouldn’t be cited on its own.
Wow, what? Are you just inventing retractions to things when you don't even know what's being referenced?
I'm referring to their textbook, widely considered the standard reference text on statistical inference.
Again, "test statistic" is just a description of any statistic that you use in a test. There's no other property required. The mean, a coefficient, a t-value, an F-value -- use any of these in a test and they become a test statistic. The fact I've had to explain this simple thing about 25,000 times is a real testament to the molasses-like slowness of these people.
It doesn't become a test because you call it so. A test is something that has the power of testing.
This is so wrong I can't tell if it's trolling or not. But if you're serious, no. Just amazingly wrong. Please just read Casella & Berger or anything and send your apologies.So if I compare my t-statistic to the current dew point in Manhattan, is that a test because I call it a test and I'm comparing two numbers?
Yup. It's a bad test, but it meets the definition of "test," according to anyone who knows what they're talking about.
Again, "test statistic" is just a description of any statistic that you use in a test. There's no other property required. The mean, a coefficient, a t-value, an F-value -- use any of these in a test and they become a test statistic. The fact I've had to explain this simple thing about 25,000 times is a real testament to the molasses-like slowness of these people.
It doesn't become a test because you call it so. A test is something that has the power of testing.
This is so wrong I can't tell if it's trolling or not. But if you're serious, no. Just amazingly wrong. Please just read Casella & Berger or anything and send your apologies.
So if I compare my t-statistic to the current dew point in Manhattan, is that a test because I call it a test and I'm comparing two numbers?Yup. It's a bad test, but it meets the definition of "test," according to anyone who knows what they're talking about.
It's not a test. The test doesn't exist.
Again, "test statistic" is just a description of any statistic that you use in a test. There's no other property required. The mean, a coefficient, a t-value, an F-value -- use any of these in a test and they become a test statistic. The fact I've had to explain this simple thing about 25,000 times is a real testament to the molasses-like slowness of these people.
It doesn't become a test because you call it so. A test is something that has the power of testing.
This is so wrong I can't tell if it's trolling or not. But if you're serious, no. Just amazingly wrong. Please just read Casella & Berger or anything and send your apologies.
So if I compare my t-statistic to the current dew point in Manhattan, is that a test because I call it a test and I'm comparing two numbers?
Yup. It's a bad test, but it meets the definition of "test," according to anyone who knows what they're talking about.It's not a test. The test doesn't exist.
Again, and please stop ignoring the point, C&B literally introduce the concept of test statistic by saying take any statistic and compare it to 5. That's using it as a test statistic. If your understanding of the term doesn't accommodate their point, then you're wrong.
Again, "test statistic" is just a description of any statistic that you use in a test. There's no other property required. The mean, a coefficient, a t-value, an F-value -- use any of these in a test and they become a test statistic. The fact I've had to explain this simple thing about 25,000 times is a real testament to the molasses-like slowness of these people.
It doesn't become a test because you call it so. A test is something that has the power of testing.
This is so wrong I can't tell if it's trolling or not. But if you're serious, no. Just amazingly wrong. Please just read Casella & Berger or anything and send your apologies.
So if I compare my t-statistic to the current dew point in Manhattan, is that a test because I call it a test and I'm comparing two numbers?
Yup. It's a bad test, but it meets the definition of "test," according to anyone who knows what they're talking about.It's not a test. The test doesn't exist.
The problem here is the notion of “test.” To say that you are testing something merely means that you are subjecting it to examination. We lost sight of this when we adopted so much jargon from economics. We aren’t evaluating these numbers, we are simply inspecting them.
Again, "test statistic" is just a description of any statistic that you use in a test. There's no other property required. The mean, a coefficient, a t-value, an F-value -- use any of these in a test and they become a test statistic. The fact I've had to explain this simple thing about 25,000 times is a real testament to the molasses-like slowness of these people.
It doesn't become a test because you call it so. A test is something that has the power of testing.
This is so wrong I can't tell if it's trolling or not. But if you're serious, no. Just amazingly wrong. Please just read Casella & Berger or anything and send your apologies.
So if I compare my t-statistic to the current dew point in Manhattan, is that a test because I call it a test and I'm comparing two numbers?
Yup. It's a bad test, but it meets the definition of "test," according to anyone who knows what they're talking about.
It's not a test. The test doesn't exist.The problem here is the notion of “test.” To say that you are testing something merely means that you are subjecting it to examination. We lost sight of this when we adopted so much jargon from economics. We aren’t evaluating these numbers, we are simply inspecting them.
Not sure exactly what you're driving at, but the point is that "test" as defined by statistics is extremely general. Yes, it's any comparison, no matter how stupid. People are confusing this with certain properties of the tests they're familiar with but don't get the root concepts.
It's not a test. The test doesn't exist.
“Testing” is not possible. We can observe, we can appreciate, and perhaps consider. But no numerical evaluation can be truly “tested” with another numerical process. That would be like asking a dog to walk a dog.
Eh, that just depends on what you mean by testing. What stats can do in "good" tests is tell you how unlikely a result would be given a set of assumptions. Maybe useful to you, maybe not.
It's not a test. The test doesn't exist.
“Testing” is not possible. We can observe, we can appreciate, and perhaps consider. But no numerical evaluation can be truly “tested” with another numerical process. That would be like asking a dog to walk a dog.Eh, that just depends on what you mean by testing. What stats can do in "good" tests is tell you how unlikely a result would be given a set of assumptions. Maybe useful to you, maybe not.
Ok, cool!
For those interested: From Casella & Berger (p. 374), a hypothesis test is defined as any rule that translates from a sample to the acceptance or rejection of a hypothesis (about a population parameter).
Then they write: "Typically, a hypothesis test is specified in terms of a *test statistic* W, a function of the sample. For example, a test might specify that H0 is to be rejected if X, the sample mean, is greater than 3."
Game, set, match.
Which immediately contradicts your previous claim that a test statistic is "any statistic that you test." They clearly state that you test a hypothesis, not a statistic.
Notice also that they never claim that the mean itself is the test statistic. Indeed, it couldn't be, since it doesn't have a known exact distribution. If it did, you wouldn't need to estimate it in the first place. So you need to find a quantity that is a function of the mean (or whatever your parameter of interest is), but whose distribution is known -- that's your test statistic, which you compare to quantiles of that known distribution to determine whether or not you can reject the null.
For those interested: From Casella & Berger (p. 374), a hypothesis test is defined as any rule that translates from a sample to the acceptance or rejection of a hypothesis (about a population parameter).
Then they write: "Typically, a hypothesis test is specified in terms of a *test statistic* W, a function of the sample. For example, a test might specify that H0 is to be rejected if X, the sample mean, is greater than 3."
Game, set, match.