Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

RE: st: RE: percent format

From   Fred Wolfe <>
Subject   RE: st: RE: percent format
Date   Tue, 14 Jun 2005 04:32:58 -0500

In case it may seem that few want the Stata additions that Jeph Herrin suggests, I want add my voice for the need for all of them, and for exactly the reasons he suggests.

Fred Wolfe

At 08:41 PM 6/13/2005, you wrote:

Thanks for confirming that there's no natural
way to format percentages. While labels work
fine if you know ahead of time what your values
are, more general applications require (as far
as I can tell) embedding (100 times) the value
in a string.

I agree that units are best stated only once,
but there are always exceptions. Some journals
have awkward guidelines for their tables and
figures. The problem most recently arose for me
in creating a series of figures where the units
turn out to be in hundredths of a percent; putting
0.01,0.02, etc along the axis with "percentage points"
in the title seems to confuse more than clarify -
practiced readers still want to see them as 1%, 2%,
etc. So I wanted 0.01%, 0.02%,... along the axis.
However, I will be making a lot of these, and sometimes
the labels will even be in half hundredths - 0.015%, etc -
so using value labels on the fly is a real nusiance.

Obviously - judging from the posts here - reasonable people
differ on the need for a % format, but that would be reason
enough to add it and let reasonable people decide what works
best for them.

On a similar note, as regards:

(although vibratory patterns in graphs are
banned until the end of time).
I often submit figures to journals, including high-impact
medical journals, that refuse grey-scale figures. As long
as Stata continues to ban hatching or speckling or such,
I have no choice but to export my data to, eg, Excel for
creating figures for submission. And while some will say
that proper labeling would get around this, I retort that
if it were so there would be no reason for Stata to support


I doubt you're missing anything. If you
say within a -*label()- option
.375 "37.5%"
you are saying that even though the
number is really .375 you want a display
of "37.5%". This is a self-knitted percent
format on the fly.
As you are being straight with Stata,
and Stata understands you exactly,
and this is well understood as a need and well
documented, I fail to see that any
trickery is involved, let alone required.
The only issue, as has been pointed out more
than once already, is that a percent
format would might less typing whenever
you want this. Those who prefer to say
"%" once can already do it concisely by
just displaying the text character; those
who prefer that a series of percents should
all bear the percent sign will miss a percent
-format- much more.
Fred Wolfe

It isn't so much the labels as the formatting of numbers.
Suppose one has a
0/1 variables indicating positive/negative. Then in hbar, for
example, I
can get .375 to display with blabel. However I really want 37.5 or
sometimes 37.5%. I can only do that by tricking Stata -
unless I am missing

At 12:36 PM 6/13/2005, you wrote:
>I have suggested this as a possibility privately
>to StataCorp, although I have two moods about it.
>In either a graph or a table, I find repeated
>percent symbols to be a constant element that
>should be factored out, as it were, and used
>just once, for example in a row or column heading or
>in an axis title.
>However, it is arguable whether StataCorp
>should play arbiter of elegance (for which
>the historical precedents are not attractive)
>whenever enough users really want something
>(although vibratory patterns in graphs are
>banned until the end of time).
>However, I don't know exactly what Jeph means here by
>conversion to string. You can use labels on
>the fly, as in
>... , yla(0 "0%" 10 "10%") ...
>the only problem being the tedium of typing.
>I'll add -prefix()- and -postfix()- options to
>my -mylabels- if I sense enough interest.
>Jeph Herrin
> > I often work with percentages, and find it somewhat annoying that
> > I have to convert numbers to strings to get them to show up with
> > a %-sign. This is particularly a nusiance with graphs, where
> > a histogram
> > that might naturally use %s on the y-axis cannot be persuaded
> > to do so without generating the frequencies first, relabeling them
> > as strings with %s, and then making a two-plot. Is there
a trick that
> > I don't know about? Is there a reason that Stata avoids, eg,
> > -format %6.1p-?
> > Seems like it would be a trivial modification at some level.
> >
> > BTW, I use Stata 9, in case it matters.
* For searches and help try:
*   For searches and help try:

Fred Wolfe
National Data Bank for Rheumatic Diseases
Wichita, Kansas
Tel (316) 263-2125     Fax (316) 263-0761

*   For searches and help try:

© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index