Re: Cluster analysis for beginners

illywhacker wrote:
On Mar 30, 6:39 pm, Jerry Dallal <gdal...@xxxxxxxxxxxxxxxxxxxx> wrote:
illywhacker wrote:
On Mar 30, 1:52 pm, Jerry Dallal <gdal...@xxxxxxxxxxxxxxxxxxxx> wrote:
illywhacker wrote:
On Mar 30, 1:36 am, Jerry Dallal <gdal...@xxxxxxxxxxxxxxxxxxxx> wrote:
Asking whether the clusters are "significant" is too vague to answer. I
suspect what the OP meant was whether the clusters are "remarkable".
This is a joke too right? You are replacing one undefined word with
another. This is indeed both remarkable and significant.
"Significant" is a word that carries baggage. "Remarkable", "extremee",
"unexpected" and a host of others get the point across better. My take
is that the OP is asking whether the observed clusters can be shown to
be "remarkable", "extreme", "unexpected". In order to do that, one must
be prepared to say what they are "remarkable", "extreme", or
"unexpected" with respect to. One could stick with "significant", but
that is more likely to raise the issues of "statistical significance"
and "practical significance", which are best set aside for the moment.
Not only do you have to say with respect to what they are
'remarkable', you have to say what 'remarkable' means! In fact, there
seems no need to introduce such a word if one is not going to define
it. If one is going to define it, then it is better replaced with its
definition: e.g. under such-and-such model, the probability of the
data is less than epsilon, or whatever. No wonder people get confused
about statistics.
illywhacker;
If you reread my initial post, you will see that it ended with
"That is, construct a measure of "remarkableness", generate a bunch of
samples that follow the "compared to" distribution, assess the
distribution of "remarkableness"."

So in effect you said nothing, since remarkableness could mean almost
anything.

illywhacker;

It could. That's why I left it to the OP.

I'm not sure what the sticking point is. The OP noticed "clusters". In order to decide if they were significant/remarkable/unexpected/noteworthy, the OP must describe the sense in which they are "significant/remarkable/unexpected/noteworthy". This will usually be relative to what is not significant/unremarkable/expected/not worth noting.

The OP asked if clusters were "significant". I pointed out that the *OP* had to define it, but to ask without a definition of what determines significance is a question that can't be answered.
.

Relevant Pages

  • Re: Cluster analysis for beginners
    ... suspect what the OP meant was whether the clusters are "remarkable". ... You are replacing one undefined word with ... "That is, construct a measure of "remarkableness", generate a bunch of samples that follow the "compared to" distribution, assess the distribution of "remarkableness"." ...
    (sci.stat.math)
  • Re: Cluster analysis for beginners
    ... suspect what the OP meant was whether the clusters are "remarkable". ... You are replacing one undefined word with ... So in effect you said nothing, since remarkableness could mean almost ...
    (sci.stat.math)
  • Re: Cluster analysis for beginners
    ... suspect what the OP meant was whether the clusters are "remarkable". ... You are replacing one undefined word with ... "That is, construct a measure of "remarkableness", generate a bunch of ... samples that follow the "compared to" distribution, ...
    (sci.stat.math)