Cold Reading Demos at TAM2

BillHoyt said:

Hey, you know what, each reviewer also has no knowledge of who got what treatment! Also, the editor who decides to publish has no knowledge. Hey, with three reviewers and one editor we're up to a septuple-blind protocol! Woo-woo!


Only the main people involved (experimenter(s), experimentees, and evaluator(s)) are the ones that are used in the terminology for blinding. You know that.


When will you stop playing churlish tr'oll games?

Bill will be playing the role of the pot today...
 
CFLarsen said:

How can he unconsciously choose one that gives better results for one group?


He could choose one consciously or unconsciously that reflects his biases. He'd go with the test or model or analysis strategy that puts coffee drinkers in the best possible light.


How will his bias not show up when his analysis is checked by someone else?


Maybe, maybe not.


How can a statistician be "less critical"? He merely makes the computations, he does not interpret the numbers.

Statisticians can be involved in every stage of an experiment, from planning and design, to analysis, to interpretation. I speak from some experience here as I consulted for a while as a graduate student.
 
T'ai Chi said:
Only the main people involved (experimenter(s), experimentees, and evaluator(s)) are the ones that are used in the terminology for blinding. You know that.


Which is why, as I said from the beginning, the statistician does not have an effect on the blinding. You are trying now for the No True Scotsman maneuver. You began with the statistician simply calculating the statistics. Now you are slipping him into the evaluator's role.
 
T'ai Chi said:
He could choose one consciously or unconsciously that reflects his biases. He'd go with the test or model or analysis strategy that puts coffee drinkers in the best possible light.

And he would immediately invalidate the experimental protocol the mament he chooses a test or model a posteriori. Read your own d*** sig line, tr'oll.
 
In case you suddenly come down with Ian syndrome:

"To call in the statistician after the experiment is done may be no more than asking him to perform a postmortem examination: he may be able to say what the experiment died of."

Recognize it? What do you suppose is the full meaning of the statement?
 
*chirp* *chirp*

T'ai,

If you are searching the net, may I suggest the terms "data monitoring committee" and that you also dive further into "response variables."
 
T'ai Chi said:
He could choose one consciously or unconsciously that reflects his biases. He'd go with the test or model or analysis strategy that puts coffee drinkers in the best possible light.

Yes, but how? You are merely repeating your claim here. You don't explain how he can unconsciously (of course he can do it consciously, that's simply fraud) choose one that reflects his bias.

How, exactly, does he do that?

T'ai Chi said:
Maybe, maybe not.

Whaa?? Are you saying that a statisticians work cannot be checked? Then, what is the use of blinding him??

T'ai Chi said:
Statisticians can be involved in every stage of an experiment, from planning and design, to analysis, to interpretation. I speak from some experience here as I consulted for a while as a graduate student.

That is not the case here. Here, the statistician comes in where the data are analyzed. In your example.

Try again:

  • How can he unconsciously choose one that gives better results for one group?
  • Is it really your opinion that a statisticians work cannot be checked?
  • If his work cannot be checked, then what is the use of him being blinded?
  • How can a statistician - in your example - be "less critical"? He merely makes the computations, he does not interpret the numbers.

Why don't you just answer the questions, T'ai Chi?
 
T'ai fly when you're having fun

T'ai fly?
Why T'ai fly?
Look to ground, look at sky,
Where e'er look I,
I see no T'ai!
Why, T'ai, why?
 
Re: T'ai fly when you're having fun

BillHoyt said:
T'ai fly?
Why T'ai fly?
Look to ground, look at sky,
Where e'er look I,
I see no T'ai!
Why, T'ai, why?


Why, Bill, why?
I have to muse.
Why try fly
preposterous apostrophe
abuse?
 
BillHoyt said:

Which is why, as I said from the beginning, the statistician does not have an effect on the blinding. You are trying now for the No True Scotsman maneuver. You began with the statistician simply calculating the statistics. Now you are slipping him into the evaluator's role.

Your appeal to fallacies is in itself, fallacious. Statisticians are well-known to do much more than just crunch numbers.
 
BillHoyt said:

And he would immediately invalidate the experimental protocol the mament he chooses a test or model a posteriori. Read your own d*** sig line, tr'oll.

Possibly. It all depends on the specifics of the experiment. Even if the model is fixed, one could still use different methods of analysis to anaylze the data.

Also, obviously the statisticians commentary on the analysis could be biased if not triple blinded.
 
BillHoyt said:
In case you suddenly come down with Ian syndrome:

"To call in the statistician after the experiment is done may be no more than asking him to perform a postmortem examination: he may be able to say what the experiment died of."

Recognize it? What do you suppose is the full meaning of the statement?

Yes, I recognize it, it is Fisher's statement about the necessary-ness of statisticians in every or most every steps of experiments, from planning and design to the final analysis. This is an ideal however, and his comment reflects that in most cases it does not take place in real life for various reasons.
 
T'ai Chi said:


Your appeal to fallacies is in itself, fallacious. Statisticians are well-known to do much more than just crunch numbers.

Not what you said originally, sir. Shall I quote you back to you again? There you described a statistician qua statistician. You furthermore misdescribed said statistician as choosing models, tests, significance cut-offs, etc. a posteriori. You are still shifting to salvage your mistakes.
 
BillHoyt said:
*chirp* *chirp*

T'ai,

If you are searching the net, may I suggest the terms "data monitoring committee" and that you also dive further into "response variables."

I don't answer in less than a day and look what you do. :) I happen to go to a thing called work.

Searching the net for answers? Oh Bill, I hope you're not foolish enough to hint or claim that I'm doing that again like you did last time.
 
T'ai Chi said:


Yes, I recognize it, it is Fisher's statement about the necessary-ness of statisticians in every or most every steps of experiments, from planning and design to the final analysis.

Really? Why? You claim the statistician has the option to switch models, tests and cut-offs after the experiment.
 
CFLarsen said:

Yes, but how? You are merely repeating your claim here. You don't explain how he can unconsciously (of course he can do it consciously, that's simply fraud) choose one that reflects his bias.


LOL, it depends on specifics of the experiment. Because there are many ways to analyze data, he could choose different model, a different analysis strategy, different software, different graphical methods, and/or be 'nice' in his commentary on the analysis. The specifics of these, of course, depend on the specifics of the experiment and the data structure.

I also didn't say that a statistician's worked cannot be checked as you try to imply. In response to your question "How will his bias not show up when his analysis is checked by someone else?" I said: "Maybe, maybe not." People could just check the calculations, people could already be consciously or unconsciously biased themselves, and a whole host of other possibilities.


How can he unconsciously choose one that gives better results for one group?


See above.


If his work cannot be checked, then what is the use of him being blinded?


See above.


How can a statistician - in your example - be "less critical"? He merely makes the computations, he does not interpret the numbers.


It was already mentioned that my example was to show how group membership can be blinded from the evaluator. I really can keep pasting this all day Claus.


Why don't you just answer the questions, T'ai Chi?

See above.
 
BillHoyt said:

Not what you said originally, sir. Shall I quote you back to you again? There you described a statistician qua statistician. You furthermore misdescribed said statistician as choosing models, tests, significance cut-offs, etc. a posteriori. You are still shifting to salvage your mistakes.

Nope. Statisticians do a lot more than crunch numbers. Despite beliefs otherwise, there is a lot of interpretation that can occur. For example, even if some things are chosen before the experiment even begins, some things might not be. What comes to mind right away are graphical methods, and most obviously, commentary upon the analysis.
 
BillHoyt said:

Really? Why? You claim the statistician has the option to switch models, tests and cut-offs after the experiment.

I merely claim that the statistican has a lot of things at his disposal for analyzing data, and as with any person, conscious, but mostly unconscious bias can creep in, which is why triple blinding seems to be a pretty good idea.
 
T'ai Chi said:


I merely claim that the statistican has a lot of things at his disposal for analyzing data.

You claimed the statistician could make those decisions after the data are collected. This violates one of the cardinal rules of experimental design. You know that, right?
 

ISF - Join now!

Every member here is approved by hand. No bots, no spam, just people who care about evidence and honest debate.

Membership is free!

Create your free account

Back
Top Bottom