Barry W Brown wrote:
Although the supposed discussion of this topic in this group seems to
have degenerated into vituperation
in which I have no desire to participate, I thought the following
observation from a talk by John Tukey at
a talk given in Houston years ago might be of interest.

I had been bothered by the fact that the null hypothesis is usually a
point hypothesis and so cannot be true

Why cannot it be true ?

(to 10 decimal places, 20 decimal places, etc).  Tukey's comment is
paraphrashed as follows.

We know a priori that the null hypothesis is false.  

Why ?

 > However, we do

not have enough evidence to say with
any certitude in which direction it is false.

Did Tukey really mean or say that ?

If I understood Neyman and Pearson correctly, allowing the point
hypothesis to be true is the foundation of the classical hypothesis
testing method. The whole idea is that whatever I say is true, unless
you have good evidence against it. It is similar to situation in court,
where what I say is assumed to be true.

We assume that H0 is true and develop mathematical machinery, formulas,
test statistics etc, assuming just that. If it then leads to unlikely
results, e.g. very low probabilities of occurring the events which have
been observed we have good grounds to reject H0.

We do not know what is the real state of nature, and accept whatever
"sounds OK" or is proposed as H0, we regard it as acceptable until
someone gives evidence against it. This means that if H0: mu=10 is OK
(true?) it may well be that at the same time H0: m=20 is also acceptable
(true?). Why? Because we never know what is the real state of nature and
null hypothesis is just a hypothesis. It is "true" only in that sense,
not as an absolute truth.

If someone doesn't like it this way it means that something is wrong
with this hypothesis testing method and one should be looking for
alternative setup.

JoJo

Although the supposed discussion of this topic in this group seems to
have degenerated into vituperation
in which I have no desire to participate, I thought the following
observation from a talk by John Tukey at
a talk given in Houston years ago might be of interest.

I had been bothered by the fact that the null hypothesis is usually a
point hypothesis and so cannot be true

(to 10 decimal places, 20 decimal places, etc). Tukey's comment is
paraphrashed as follows.

We know a priori that the null hypothesis is false.

However, we do
not have enough evidence to say with
any certitude in which direction it is false.

Although the supposed discussion of this topic in this group seems to
have degenerated into vituperation
in which I have no desire to participate, I thought the following
observation from a talk by John Tukey at
a talk given in Houston years ago might be of interest.

I had been bothered by the fact that the null hypothesis is usually a
point hypothesis and so cannot be true
(to 10 decimal places, 20 decimal places, etc). Tukey's comment is
paraphrashed as follows.

We know a priori that the null hypothesis is false. However, we do
not have enough evidence to say with
any certitude in which direction it is false.