The problem must be quite challenging, because despite the interest (even a
thread in this newsgroup popping up on the topic every few months), there
seems to be no solution; but, actually, someone did it!
It's in Japanese - the author apears to be a real fanatic of both statistics
and Excel (particularly charting) - kinda like myself
After literally
weeks of searching the web on this very topic (just to challenge many people
claiming it couldn't be done; otherwise as a biostatistician I do survavival
analysis literally on daily basis with statistical software packages) a few
months ago I found this great site - a real jewel:
http://aoki2.si.gunma-u.ac.jp/lecture/stats-by-excel/vba/index.html
which apears to be a really professional statistics course page
(introductory/intermediate level, probably for nonmathematicians) with an
impressive collection of statistical methods implemented in Excel (including
quite a few multivariate methods and some less common methods for
categorical data), while having full awareness of and links to the studies
of Excel's known bugs/inacuracies in some statistical functions.
Many times I've wished someone translated that page and all its software
(dozens of workbooks with macros) into English - I may be fluent in several
European languages, but unfortunately I don't know Japanese - actually,
judgings by the names of the html documents and Excel files belonging to the
website the author (S. Aoki - maybe a statistics professor?) knows English,
so perhaps eventually ...
Anyhow, the link no. 22 in the second section (don't ask me what the names
of the two secions are - judging by the contents they're not univariate and
multivariate, or description and inference - but it doesn't matter) is, yes,
Kaplan-Maier survival curve (it's actually one of the few words written in
latin characters, so its easy to spot). And at the top of the page the link
takes you to is a link to the Excel worksheet (with VBA macro) that draws
surprisingly pretty K-M survival curves (it also estimates the curve with
properly computing standard errors of survival probabilities, which is not
so simple).
Of course, if you want no Japanese characters on the plot (and also taylor a
bit the fonts and the like) you have to mess with the VBA - I'm really an
amateur in this field but with some luck I managed it and it works
perfectly. But I don't have the file on this computer, and anyway I don't
feel comfortable with sending it to anyone because I didn't ask the author
for permission to modify the code, while I really admire anyone who dares
using Excel in the professional statistical community (yes, I happen to be
one of those black sheep), which is so full of mathemathicians with
religious worship of S-Plus for analysis and Tex for publishing.
Speaking of that topic - the marketing research & SPSS people have more
simpathy for Excel - apparently like the poster, because his email address
says that he's from the company representing and selling SPSS in Slovenia.
So, I'm happy to post this for three reasons - to praise Mr. Aoki's site, to
be able to help in a newsgroup thread where such authorities as Mr. Peltier
and Mr. Mehta posted, and to help a compatriote (there is no surname in the
e-mail address, so I can't tell if I personally know him - Slovenia is tiny,
its capital Ljubljana even more so, and the statistical community within it
even more so).
Gay Vidmar
University of Ljubljana, Institute of Biomedical Informatics
