DF test at 5% -- your "power" is
50% at that value of N. That is, on replication, you expect to do
better or worse that this, even odds.
For a 1-df chi squared, doubling that N is what is needed to increase
the power to 85%.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpili
d, subtracting 200, if you want to test those means.]
Similarly create a variable that has 192 means, one for each vignette,
and use that as covariate for ANOVA by 200 IDs, if you want those as
adjusted means.
I don't know what econometricians encounter.
--
likely
of these choices, if the program is a standard).
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
===
This list is open to everyone. Occasionally, less thoughtful
people send inappropriate m
ranks to normal-deviates. Whichever
of these you do, the main criticism you receive should be about your
data collection which left these options.
The far better technique is to ponder and figure out your
scaling/scoring questions before you collect the data.
What did you ho
on how many outliers there are. Check on how the
"significance" of the correlation depends on the DF.
This is not precise, but I maybe it helps ...
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
===
ons and structure common in clinical research and
many other places. With dichotomous measures, you might need more,
since the correlations are attenuated. With unambiguous structure,
you would need less.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pit
the parts meant - 1897 and 1899. And Yule recognized that the
data did not need to be bivariate Normal.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
===
This list is open to everyone. Occa
pt that
it probably doubles or triples the sample size needed for stable
correlations for PCA, for the reasons I just said.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
===
This list is
were reported in a misleading
way.
I can't say what you are comparing, but I think you want to compare
directly instead of trying to do adjustments.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
=
t's bold. There is not much that has been used often, that really
makes use of MV settings except for assuming MV normal, and using
Mahanalobis Distance. - that I know of.
Well, you have to place some faith in your metric for each variable,
regardless. Do a nearest-neighbor algorithm, and lo
On Tue, 23 May 2000 13:49:38 -0700, "G. Anthony Reina" <[EMAIL PROTECTED]>
wrote:
> Rich Ulrich wrote:
me >
> > > Plus, I've run the multiple
> > > regression without the transform and se
ability, or the interesting "variance."
But if you are trying to compare "variance in the dependent variable"
when that variable is measured in Counts, then you HAVE to measure it
in Counts.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
=
tly describe the tests, they can test variability starting from
that same principle: start with the difference between each score and
the Mean. Or median. Then you can use Absolute value (Levene) or
squared value or log of the square value.
--
Rich Ulrich, [EMAIL PROTECTED]
http:/
> Engineering Statistics, 2nd Edition; Bowker and Lieberman, Prentice Hall
> Introduction to Statistical Analysis, 4th Edition, Dixon and Massey,
> McGraw Hill
< snip >
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
=
bandon some of the post/response format.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
===
This list is open to everyone. Occasionally, less thoughtful
people send inappropriate messages. Please D
stion might be, "How much precision is
lost, or new error is introduced, when replacing the old with the
new?"
Or are you trying to predict the other *measure* with least-squared
error? - then the regression line meets that criterion.
--
ffended if you called up and asked them to take a
personality test on the phone, or an IQ test, than if you ask their
opinions about the president or Congress.
But if attitudes are malleable, the measure of personality had to go
beyond a simple summation of attitudes, so the previous focus had to
ul? In an exploratory mode, you should look at them
all.
With your data rescored as (positive) Difference, you can compute
tests within data sets, and between data sets. It sounds as if the
ultimate comparison Between data sets might properly u
of the studies, ever
done by anybody, had ever controlled for anything. This was in the
local newspaper. I keep pretty flexible in my expectations for the
local newspaper.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
=
on't know what your comment means, " ... without taking into account
their respective means."
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
===
This list is open to everyone. Occasi
elves on the
scientific-cutting edge, having read the BJM. Of course, the
non-funding of studies with N *over* 300 is ever-justified because
the studies would be too difficult, and/or would cost too much.
Okay, Steve, that was my first response...
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu
to have the government
take up the slack.
(There's one traditional distinction between the parties in U.S.
politics -- "Big-money" regularly buys or rents the support of the
Democratic party, whereas the Republican par
rue is THAT? - concerning the original, or the critique."
Anyway, "misused stats" like a bad F-test are not very interesting. I
have seen an illegal transformation; I have seen a t-test of less than
1.0 that was mistaken for "significant". These were never published
tical, mathematical drawbacks?
Have I covered it? There are also comments in my stats-FAQ
and in some other on-line descriptions of regression.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
===
T
ion; in any case,
adding-and-subtracting those high-loading items is not
apt to make a very robust total score. I take this as a warning
that I may have too many variables for the N, and I should
test some smaller subsets. Having more than one or two
items with "split-loadings" bothers me
re most likely to be interested
in; so there's a couple of reasons for you.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
===
This list is open to everyone. Occasionally, less thoughtful
people
of
tho old ones; so I don't need GUI for much, even when I use my PC
version.
Mostly, I still use SPSS 6.x on VAX, since that lets me readily edit
and shorten the output file, using any of three or more different file
editors, while I am deciding what to pri
it.
You use Roy's if you think there is a simple effect and (for some
reason) you can't describe that in advance.
--
Rich Ulrich
http://www.pitt.edu/~wpilib/index.html
===
This list is open to everyone. Occasiona
> >
> > The distribution of the quantity I am looking at could be approximated by an
> > gaussian (just in case it eases the discussion). At least it looks like a
> > gaussian, when I histogram it.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
you reproduce your present data, with dangerous assumptions?
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
===
This list is open to everyone. Occasionally, less thoughtful
people send inappropria
has a reference on ICC for an unbalanced design. It
entails approximations, so I hope the design is not *too* unbalanced.
< snip, McGrath sig.>
< snip, Bob Wheeler post; included for no imaginable reason. >
< snip, quoting of Edstat-L message from the bottom of Bob Wheeler's
er, the winner
in the long run is the player with deeper pockets at the
start.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
===
This list is open to everyone. Occasionally, less thoughtful
people sen
dd values
that you might have because of serial correlation, then you can accept
this set as robustly Normal, in regards to whatever you are testing.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
re arbitrarily written as positive -- by multiplying loadings
by minus one when the program detects more signs in the "wrong"
direction). If you look at the items and you can't figure why the
question is in the "opposite" direction, then maybe you have trouble
with your
chastic temperature effects. The value I would be interessted in, is the
> _length_variation_ and an _error_estimation_ for this value.
>
> The distribution of the quantity I am looking at could be approximated by an
> gaussian (just in case it eases the discussion). At least it looks lik
- typographical correction -
On Tue, 25 Apr 2000 10:26:02 -0400, Rich Ulrich <[EMAIL PROTECTED]>
wrote:
> On 24 Apr 2000 07:41:10 -0700, [EMAIL PROTECTED] (Paul W.
> Jeffries) wrote:
>
> > Does any one know of good sources--either book chapters or articles--tha
) 2nd edition.
If yours is very narrow question, the answer is that "attentuation" is
direct and estimable, and you can look for that word in your table of
contents or index.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pit
> On 21 Apr 2000 17:25:01 GMT, [EMAIL PROTECTED] (Wen-Feng
> Hsiao) wrote:
>
> > Dear Ulrich,
> >
> > Thanks for your reply. I have obtained a book of Agresti from library --
> > Statistical Methods for the Social Sciences, 2nd edition. I fortunately
> > locate Section 7.4 titled 'comments con
,5)
in the alcohol data.
I have never seen that text, but my local library supposedly has it,
so I will try to say more, after I have seen it.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
===
Th
biggest, because most critters never stop growing. That seemed to
tie in to the rat-life-spans, too.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
===
This list is open to everyone. Occasio
3), then the Confidence Interval is all above 3.0.
> what confidence do we have that the treatment effect is AT LEAST 3 lbs?
It depends how much of the CI is above 3.0, doesn't it?
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
===
> don't seem to help in this issue.
I think you want to compare judgments rather than comparing Kappas,
but you need to define a purpose. In what fashion is something
expected to be better or worse?
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
=
I know--and I have asked about this
before. So, statisticians have not considered the question especially
noteworthy. (And, I am curious, Do recent Experimental Design texts
say anything?)
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
=
f these as small or medium-to-large effect sizes?
Go back and check Cohen and I think you will see that he was careful
not to overgeneralize about r and d . His levels seem to have
*little* to do with your debate, since your studies are a-typical.
He is talking about "the usual studies&quo
orted into groups based on (X+Y) answers, you don't really
have a fair hypothesis, to ask about the presence of (X) and (Y)
separately; but you should tell about them, as descriptors.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
=
way, the standardized item-alpha (provided, for instance, by
SPSS Reliability procedure) is computed STRICTLY from the
correlations, with no reference at all to variances. Usually, that
should be an optimistic estimate.
Paul Gardner laid out most of the relevant questions, quite well.
--
Ric
but one of the best organizing principles that we have -- right
now -- is that of framing questions as hypotheses.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
===
This list is open to everyon
( how did we get to HERE, from Data Mining?)
On 15 Apr 2000 17:50:05 GMT, [EMAIL PROTECTED] (Radford Neal)
wrote:
> In article <[EMAIL PROTECTED]>,
> Rich Ulrich <[EMAIL PROTECTED]> wrote:
>
> >One thing that remains true about stock investment schemes: There ma
u can work around the edges, and try to figure what stocks
really *ought* to have been the ones in that group, before eager
anticipation drove their prices up.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
===
aped curve" has been dated to Jouffret, 1872 -- who
was (particularly) describing the bivariate normal.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
===
This list is open to everyone. Occasi
med after the person who first
discovered or discussed it. Stigler goes on to observe that, more
often than not, the famous person had very little to do with that
particular observation.)
Until someone has another reference, I think the question is still
open
--
Rich Ulrich,
ing, you can probably be quite safe in doing the
two-way ANOVA -- safer, I suspect, than anything you can do with
ranking as the first step.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
==
quot;Criticism and the Growth of Knowledge" (Lakatos, ed.). This book
happens to be from the proceedings of a symposium devoted to exploring
Thomas Kuhn's thesis about revolutions in scientific discovery, and it
is a modern classic.
--
Rich Ulrich, [EMA
rnative. The standard testing is called Two One Sided Tests, to
show that the amount is definitely less that the Upper limit, and
definitely greater than the lower. Basically, you need to construct a
Confidence interval on the difference and have
forms close to the
general test on a good range of "general" alternatives.
The last one has a subjective element, since different folks are apt
to have their own experiences that they would prefer to see weight
given to, of what the general alternatives ought to look like.
--
s, compared to any test that does treat them
as Ordered, whether you collapse them into categories or not.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
===
This list is open to everyone.
zes.
Trying to convert those two measures to something else for a
meta-analysis is like trying to convert your "dollars" for a shopping
trip to New York City -- neither rubles nor cartons of cigarettes
would be nearly so negotiable in NYC, though they serve a similar
function in other p
.
All of that is in one wave; and you hope that the items you are
tracking are not correlated with the difficulty of finding the people;
else, you might have to make estimates about *why* there is a
correlation.
Have you messed up the question?
--
Ri
ut there are
several different ways to combine results from multiple studies. Do
you elect to combine the p-levels or combine some measure of
the effect-size, and do you weight studies equally, or according to N,
or according to precision of result? - if that is different from N.
--
Rich Ulrich,
weight to the 0/1 comparison. Obviously, the
test after forcing a transform to ranks is even weaker in its weight
for the higher groups.
That is not all that I have hoped to say, but that may be all that I
get to, this time around.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/in
lp-me!
http://www.stattransfer.com/lists.html
- has all the advertised stats lists, and automates the subscription
process. You'll have to see if anything comes close enough for you.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~
If
rho approaches 1.0, then you see it edge-on, as just a single straight
line, which similarly fades out, at the extremes.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
===
This list is open
ng
> >provides an OR of 250:1. That is *one* way to say that I disagree.
> >
< snip, delete, some other topics >
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
===
This list is open
idence Interval, please hire a
consultant.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
===
This list is open to everyone. Occasionally, less thoughtful
people send inappropriate messages. Pleas
ansfer.com/lists.html
which has information about joining various mailing lists.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
===
This list is open to everyone. Occasionally, less thoughtfu
etric is
important, if you want to achieve the natural (and symmetric)
rejection region.
I have made statements on this before. I am still trying to get it
all complete and concise. I think I will have a bit to add to the
above, on Sunday.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/
iew, and
that includes similar methods under other names which date back to the
1930s.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
===
This list is open to everyone. Occasionally, less thoughtful
peo
;.
When it comes to a variable like "race" and "socio-economic
background" which are just about *inextricably* intertwined in the
U.S., racialist ignoramuses will avoid the problem by never noticing
that the confounding exists.
--
Rich Ulrich, [EMAI
O put it another way, you can't assume that the only goal is to
detect an effect as being non-zero. In fact, I think it is pretty
useless to cite 95% CI's as an "effect" when the test is barely at 5%;
the range is just LARGE.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.ed
entry of variables for logistic is just as bad an idea
as it is for OLS regression. Especially relevant for small N: Note
that logistic is even *worse* than OLS about capitalizing on chance.
You can read about stepwise, and references, in my stats-FAQ.
--
Rich Ulrich, [EMAIL PROTECTED
tress the point. But it was a
good point.
I hope that the researcher is not DEPENDING on p=0,05, in order to
"distinguish between absurdities and relevant relations".
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
==
riation observed in strata?
I don't know that I could answer it, even if it were worded
differently, but this question does not seem complete.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
==
a disqualifying test; I have to live with it, if it
happens to meet the 5% level, all by itself. So, I worry about it,
but I don't include it in my Hypothesis test -- unless it has been
elevated, beforehand, to be a separate aspect of the main hypothesis.
--
Rich Ulrich, [EMAIL PROTECTED]
ht
; is usually a bad idea. See the comments in my
stats-FAQ, and their references. (If you are worried about correcting
for multiple tests, then you probably *shouldn't* add the variable
because it is probably capit
I have never
faced regression "weighting" for real data; I would try to clean up my
data or justify it, instead of weighting. But there are other areas
(certain survey research) where the use of weighting -- for various
reasons that probably do not match yours -- is the standard
On Thu, 09 Mar 2000 20:20:47 GMT, [EMAIL PROTECTED] wrote:
< ... >
> 1 Independent Variable (Group)
> 13 Dependent Variables (various IQ and cognitive testing scores)
>
> The MANOVA was significant and I wanted to perform post hoc tests on
> Group. Here then is the caviat:
>
> GroupN
>
e the Oxford English Dictionary?
Texts written for ESL courses ("English as a Second Language").
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
===
This list is open to everyone. Occasionally,
ctor analysis across some key scores.
For accuracy: you will vastly reduce your complexity if you designate
a modal score, or a "gold standard" of a correct score. Then you will
have 16 comparisons instead of 120 (or, does order matter?)
--
Rich Ulrich, [EMA
think you have to *read*, you have to be
exposed to a large number of various situations. You have to read
some good examples, and you have to read criticisms which include
examples that were not-so-good.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
==
.." His reference for Fieller is (1940) "The
biological standardization of insulin"; Journal of the Royal
Statistical Society, Supplement, 7: 1-64.
It is rather complicated, but it simplifies a lot if the denominator
is large compared to its standard error; and it simplifies even more
if
ey want just like everyone else. And
everyone else has both the right and the means to ignore them."
=end cite
- By the way, isn't it still proper form to edit what you reply to,
and especially to delete the lines of people's .sigs ?
--
Rich
by, "beginning or end of the data
set", in this context. When the data are ordered, in a meaningful
way, then that is something that might become a factor --
statistically speaking -- in the analysis.
You might have to find someone to speak to, with your data in hand, in
order to
ot
emphasized in the question. If using that if it is not a convention,
already, in your area of specialty, then you have enough troubles
explaining why you used that regression ... without getting into the
details of *that* adjusted
However, for certain purposes, you might want to use the
mid-point of the range implied by that categorical response. For two
coin-flips, there are
0.25 responses with more heads than 1,
0.25 with fewer, and
the midpoint of the category with 1 head is 0.50 -- while
0.75 is the cumula
transformation may fall outside of that family.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
===
This list is open to everyone. Occasionally, people lacking respect
for other members of the list send
smaller cohorts where there are covariates for each individual.
Look for comparable literature in your own area.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
===
This list is open to everyone. O
iousness" sounds like heavy jargon, but offers
some success in being specific. Here is a link that leads also to
papers that you can read on the subject, if you are that serious.
http://www.emory.edu/EDUCATION/mfp/effpage.html
--
Ri
iscriminant function. And so on.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
===
This list is open to everyone. Occasionally, people lacking respect
for other members of the list send message
MANOVA; to
compute the difference and put it into a new preference-variable,
pref_a = (att_a - att_b) ;
and test that variable with a t-test.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
===
t; same data set ?
Can you get two for the price of one? No.
(How much space does your textbook want to devote to this topic?)
If the textbook explanation of the logic is not clear, you might try
the original references -- where the authors are the experts and have
more pages available.
--
Ri
in with an answer that makes you happy, could you
explain the question more slowly?
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
===
This list is open to everyone. Occasionally, people lacking
c test might test what you want to test?
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
===
This list is open to everyone. Occasionally, people lacking respect
for other members of the list se
In response to,
> > > Rich Ulrich wrote:
> > >
> > > > Prior to artificial insemination, sperm can be sorted by
> > > > weight in a centrifuge, since sperm with Y-chromosomes generally
> > > > weigh less than the X-chromosomes, by the
hin the week (typically) that your
ISP saves it, or it will be gone from them. With a little more
difficulty, though,
[GOOD] the message is *never* gone, because you can fetch it from
DejaNews or ReMarq (assuming it was not posted with X-NoArchive: Yes).
Hope this helps.
--
Rich Ulrich, [
ate those
by linear algebra to scores of {low=1.0, next-low=2.0, next_a= 2.2,
next_b=2.3, next_c=2.5} -- or whatever, to see how the scoring
compares to {1,2,3,4,5}. Agresti's textbooks give a detailed example.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.
lar form, and that is most of the *assumption* required if you
want to do something that is parametric and more flexible.
Sorry; one of the things I most dislike about "nonparametric testing"
is the over-selling of it -- "reduced distributional assumptions"
would be a more mea
is the test performed on the difference in
means.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
===
This list is open to everyone. Occasionally, people lacking respect
for other me
child after the
> > > first is substantially more likely to match the sex of the previous child
Note, "substantially". Huge datasets haven't been enough, yet.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
On 24 Jan 2000 14:37:34 -0800, [EMAIL PROTECTED] (Drake R.
Bradley) wrote:
> Rich Ulrich wrote:
>
> > On 24 Jan 2000 09:08:59 -0800, [EMAIL PROTECTED] (Bushaw, Gordon
> > -ADMIN) wrote:
> >
> > > A related question-
> > > I seem to remember reading or
one have any information about this, or is it complete
> nonsense?
It would be complete nonsense if anyone claimed that.
But you are probably imagining it.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
301 - 400 of 459 matches
Mail list logo
