Why is it time and content seem to intersect on Saturdays?

Whatever the reason, I did have a minor success with my PowerMac and R
yesterday.  I have been limping along with R on the Mac, using
only precompiled library code.  And since one package I tend to
use a lot (Hmisc by Frank Harrell and friends) didn't work right, I
have been hampered by having to export datasets to my PC workstation
and run things there that run under Hmisc.  So what was the
success?  I figured out:

1.  How to run the source code compiler for R on the Mac.
2.  I discovered that the compiler for R needed an implementation
of g77 (I know, that's not as good as g95, but it's what was required
by Hmisc).
3.  I figured out how to load up the latest g77 code.  (I really like the website hpc.sourceforge.net.)
4.  I got Hmisc compiled and running - hooray!

Of course, this sucked up about an hour and a half, but it was worth it.

Let's see ... it's been about two weeks since the last blog
entry.  A couple of people have asked me (in person) about the
lipid stuff I was working on.  It was encouraging to be reminded
that there are other "story junkies" out there, who find no peace until
they've completed a book/movie/etc.

Well, time for the first disappoint:  I don't have the answer
yet.  We did get a download of additional data that was more
detailed than the standard lab reports (thanks to the generous sharing
of the lead science officer at the company doing the lab tests). 
The results we saw using the more detailed data were provocative, but
unfortunately the 400+ number of people in the current study is just
not enough to answer the question we are asking. 

(Sidebar:  it will likely be enough to answer whether or not there
is added benefit to the people taking the drug being studied, but not
enough to answer this tangential question.) 

(Second sidebar:  In astronomy, if you can't see a particular
asterism, say the whirlpool galaxy for example, you get a bigger
telescope.  The equivalent in the statistical world is the number
of people in your study.  If you can see a difference between two
groups but the difference isn't significant, you probably need more
people in order to see the difference.)

We got to the point on 9 September, recognizing that we couldn't pin
down the effect with the current study, and so we turned our attention
to other things.  However, I did mention to the head of science at
the company providing the lab test info that, if he came across a
larger study that would let us examine the question again, let me know.

Not quite end of story.  The next Monday I get an email from my
new friend at the company that does the lab tests.  It seems he
was just given permission to use the dataset from a larger study that
his firm has been providing with lab tests.  He normally works
with a statistician friend who is at the CDC, but that fellow is
currently swamped.  So, he wondered if I would be interested in
helping him.  He has a specific target in mind and part of what I
would get out of the partnership is an opportunity to chase down the
effect I saw in a larger study - about 1600 people this time.

Obviously, there is a lot of "hand waving" going on here.  No
names, etc.  It's the nature of the beast that you don't talk
about people or companies without their permission, at least not until
the joint work appears in press (and is obvious anyway at that point) -
which tends to take months to a year.  But life goes on and you
realize that God is much more in control of the overall process of life
and research than you are.

One of the most fun parts of doing research is discovering new things
(another is working with interesting people, but I digress...). 
Recently I've been working with another fellow on analyzing and
understanding some patterns we've seen in the relationship between
coronary calcium (this is what "Heart check America" and similar
electron beam computed tomography tests - EBCT - are measuring when
they look for evidence of heart disease in otherwise seemingly healthy
people) and a variety of physical and lab measurements.  We have
some baseline data from a study we are working on (baseline = before
they start taking study drug or placebo) and have been looking at the
relationship between lipids (ie cholesterol measures) and coronary
calcium, and have come across a relationship no one else has so far
reported.

Normally, when you think you have "discovered" something new, your
knee-jerk reaction should be: "I don't think so...".  Healthy
skepticism is the sane first response.  So, we looked for all
kinds of other possible reasons (errors in coding or in data transfer,
confounding variables that create the impression of a relationship
where there is none, etc.).  At this point it looks real. 
We've requested some additional data from the people who do the test
and soon (hopefully tomorrow) we'll have a better idea whether this is
really something new --- or just something old that we mistook for
something new.

This is the kind of thing that keeps the adrenalin pumping when you sit
and stare at the screen for hours on end looking at numbers ... numbers
... numbers ...

Summer vacation is almost over. We had a great time driving around Lake Superior with Harald and Naomi Schaefer. Funny how you can be so close with friends you haven't seen in years. It was great to be able to speak German again. (Ich muss mal zugeben, dass ich meine "zweite Muttersprace" doch vermisse!)

But with work rapidly approaching (only about 10 hours away...), I got my computers fired up and am busy checking email, etc., and the latest blog entries. It's nice to be away and it's nice to come back.

Many thanks to Joel for the reading material...

About two weeks ago we learned that the state of Illinois is thinking of eliminating our farm as part of the Prairie Parkway (www.prairie-parkway.com) project. Funny how things that seem to last for ever (92 years in the family so far) are suddenly very vulnerable. Needless to say we are doing our best to rally support for the expressway to be built somewhere else.

A few weeks ago I bought a PowerMac with dual 2.0 processors.  I
had looked through the reviews and was excited about working with
it.  Along the way I had seen a review that suggested it would run
about 25% faster than a dual 3.2 Xeon from Dell, so I thought that
would be pretty cool.

I was running a particularly long analysis recently and had the chance
to compare the PowerMac to the Dell, and was surprised to find that it
ran about as fast as you would expect a 2.0 G machine to run - in other
words, quite a bit slower.

I've thought a little about what could be going on.  The same
program was running on both machines (R - for those who are familiar
with stat software).  But I was running a version on the PowerMac
that I had downloaded, rather than building from scratch.  My
suspicion is that the PowerMac software was compiled on a 32bit machine
and, hence, it fails to take full advantage of the 64bit architecture.

Why don't I build it from scratch myself?  Learning curve and
limited time at the moment.  I hope to do that in the "not so
distant" future, but time will tell.

Which doesn't change the fact that I enjoy using the mac.  Change is good.

Anyone with PowerMac experience with R is invited to comment on whether
my suspicions are true - especially if they have compiled R from
scratch for OSX Tiger.

width="350" align=center border=0 cellspacing=0 cellpadding=5>

Your #1 Match: INFP

The Idealist You are creative with a great imagination, living in your own inner world.Open minded and accepting, you strive for harmony in your important relationships.It takes a long time for people to get to know you. You are hesitant to let people get close.But once you care for someone, you do everything you can to help them grow and develop. You would make an excellent writer, psychologist, or artist.

Your #2 Match: INFJ

The Protector You live your life with integrity, originality, vision, and creativity.Independent and stubborn, you rarely stray from your vision - no matter what it is.You are an excellent listener, with almost infinite patience.You have complex, deep feelings, and you take great care to express them. You would make a great photographer, alternative medicine guru, or teacher.

I'm sitting here at my real office (not the one in Chicago, where I go
so that I can run from meeting to meeting instead of getting work
done), taking a break as I transition from one analytic project to the
next.  This morning we were figuring out how to use echo
cardiogram measures of strain in heart muscles to diagnose early stages
of hypertension, diabetes, etc.  This afternoon is "death by
temperature":  studying the relationship between ambient
temp/humidity and daily death rates in the 109 largest US cities. 
Great stuff, huh?

I'm enjoying the new computer my son helped me pick out, as well as the
new displays.  Originally I thought I would try to get a big
monitor for the new Mac, but then I discovered the value of dual
displays.  I don't watch DVDs, so there would be little benefit in
having one of those superexpensive DVD displays that Apple is famous
for.  Instead, for less than $600 I have dual 1280x1024 displays,
which are great for programming.  (I like the 17s better than the
19s because they take up less room.)  Not only that, but the
Windows workstation goes in the analog inputs, the Apple in the
digitals.  The primary Apple window is on the left, primary Win on
the right.  A press of the button on the front of the monitor
switches between the two systems, so I can either have the primary of
each or have duals for Win or Apple as needed.  I don't use the
word "sweet" - but you know what I mean.

And as I sit here, I think of my son, sitting in front of his dual
displays, typing away, too.  Like son, like father, I guess.

Xanga

I wouldn't be here if my son wasn't moving to NYC ... but then, it'll be fun to visit him.