Hi. My interest in Jaynes started years ago while getting my Ph.D. in electrical engineering. Downloading "Probability Theory: the Logic of Science" cleared up...
2
Kevin S. Van Horn
ksvanhorn
Aug 26, 2003 6:22 pm
My website for errata and commentary of PT:TLOS has been moved to http://leuther-analytics.com/jaynes/. I've also updated it; see "What's New" for details. I...
3
buybuydandavis
Aug 26, 2003 8:31 pm
In the space of the 12 hours or so since I announced this group on the MAXENT list, 17 people have joined. Buy Buy -- Dan Davis...
4
Kevin S. Van Horn
ksvanhorn
Sep 3, 2003 4:08 pm
Jaynes doesn't say much at all about how, in general, to find the parameter values that produce desired expectations in constructing a maximum-entropy prior....
5
Kevin S. Van Horn
ksvanhorn
Sep 21, 2003 12:25 am
I've just updated my PT:TLOS web site. The new information includes - additional commentary on computing the parameters of a maxent distribution - additional...
6
buybuydandavis
Oct 13, 2003 10:40 am
Last weekend I read Donald Gillies' "Philosophical Theories of Probability", 2000. While he covered the twentieth century, he unfortunately chose Keynes as the...
7
Niko Brummer
bonitofreedive
Oct 13, 2003 1:46 pm
A short introduction, and then a problem: I am an electronic engineer and I have been doing years of research and development on the inference problem that is...
8
buybuydandavis
Oct 14, 2003 2:33 am
Hi, Niko. It's not clear to me what you're trying to get at with your post. Yes, sometimes data makes you less confident in your ability to predict what will...
9
Kevin S. Van Horn
ksvanhorn
Oct 14, 2003 3:31 am
... Your constraint does not uniquely determine the distribution. Let x = wine/(wine + water); then any distribution for x that is symmetric about x = 1/2...
10
Kevin S. Van Horn
ksvanhorn
Oct 14, 2003 4:06 am
... What advantage is there to pretending to be more sure of yourself than you actually are or have any right to be? The kind of situation you're talking about...
11
Niko Brummer
bonitofreedive
Oct 14, 2003 7:33 am
... Hi Dan and Kevin I was certainly not trying to manufacture a "paradox". I was just trying to get my head around a somewhat surprising fact. (I am in a...
12
Kevin S. Van Horn
ksvanhorn
Oct 15, 2003 3:08 am
... The source of the problem here is that the term "knowledge" is misleading. Replace "knowledge" with "certainty" -- a more descriptive term -- and the...
13
Kevin S. Van Horn
ksvanhorn
Oct 15, 2003 4:18 am
... Make that, "average predictive distributions" instead of "average predictions."...
14
Niko Brummer
bonitofreedive
Oct 15, 2003 7:01 am
... Thanks. This is suggestion certainly helps. ... Sorry. I put this in the wrong way. I was trying to paraphrase the meaning of "expected"; and I didn't do it...
15
Niko Brummer
bonitofreedive
Oct 15, 2003 7:10 am
Another thought on language usage: When using the term "THE answer" I think this should denote the latter : "the answer of the oracle". The former type of...
16
Niko Brummer
bonitofreedive
Oct 15, 2003 7:35 am
Still working with yes/no questions and (Fisher) sufficient statistics and posteriors (PT:TLOS Chapter 8): Fact 1: A statistic S(data) is sufficient for this...
17
buybuydandavis
Oct 15, 2003 8:56 am
... about ... I grant that I have not demonstrated a transformation group or symmetry property which determines the prior. First, we'd have to decide whether...
18
Kevin S. Van Horn
ksvanhorn
Oct 16, 2003 1:47 am
... Yes, the Jeffreys prior seems plausible, but so does the uniform distribution over x = wine/(wine + water). No, this is not a show-stopper problem for the...
19
Kevin S. Van Horn
ksvanhorn
Oct 16, 2003 1:54 am
... I don't think this is a meaningful thing to do. A sufficient statistic is a function of the *data* only, and relates to the sampling distribution /...
20
Niko Brummer
bonitofreedive
Oct 16, 2003 8:37 am
I just found this link to the book in PDF form: http://www.stat.washington.edu/hoff/courses/stat564-2001/JaynesBook/ (I have received my physical copy of the...
21
Niko Brummer
bonitofreedive
Oct 16, 2003 9:56 am
... Thanks Kevin. Below I'll concede your point but then motivate under what circumstances the posterior IS a statistic: 1. An immediate disqualification for a...
22
Niko Brummer
bonitofreedive
Oct 20, 2003 6:37 am
I stumbled on this today: http://francisperey.com/QM/index.shtml, from which I quote: "One of the most interesting application by Jaynes[13] of his group ...
23
Niko Brummer
bonitofreedive
Oct 20, 2003 10:58 am
I found a passage in PTTLOS that shed some light on my first issue that I raised in this group. See "13.5 The Honest Weatherman". Here the situation is...
24
Kevin S. Van Horn
ksvanhorn
Oct 20, 2003 2:06 pm
... The last line is incorrect: only events that actually happen are relevant. It should instead read, "If in the extreme he gives p=0 for an event that does...
25
Niko Brummer
bonitofreedive
Oct 21, 2003 7:02 am
I read up a bit more about "proper scoring rules". See e.g.: "Inductive Inference and the Maximum Entropy Priciple" by N.C. Dalkey in Maximum Entropy and...
26
Niko Brummer
bonitofreedive
Oct 21, 2003 12:36 pm
And here is a publication where "proper scoring" is actually applied to weather forecasting: It starts with the following sentence! "Operational weather...
27
Kevin S. Van Horn
ksvanhorn
Oct 22, 2003 3:38 am
... Bernardo and Smith's book, _Bayesian Theory_, has a nice discussion of this. They also have a nice discussion of de Finetti's Theorem (section 4.3),...
28
Niko Brummer
bonitofreedive
Oct 22, 2003 7:16 am
... Thanks Kevin. I have a copy of this book on loan from our local university library. I read parts of it months ago. Then I had no interest in ...
29
Niko Brummer
bonitofreedive
Oct 23, 2003 2:58 pm
At times Jaynes's Robot vaguely irritated me, but I think I now understand why he used a robot. I have also found a very useful job for this robot. Why a...
30
buybuydandavis
Oct 24, 2003 2:09 am
... True. But given the vague nature of the question, it shouldn't be surprising that there are different plausible priors. ... More work is needed, but I...
