Help with high resolution graphs

[Denvil Duncan]

Hi All -
The editors at the Economics of Transition have requested that I
increase the resolution of my graphs to 600 dpi (dots per inch). The
graphs were created in stata, then copied and pasted into word, then
converted to pdf. The images in the PDF file should be 600dpi.

I think the change should be made at first stage (stata) but I may be
wrong. Any help with this will be appreciated. I need to submit the
new images by Wednesday 21st Oct.

Cheers

DD.

[Jason Delaney]

For art resolution issues, I'm your man, Denvil. It's hard to know
exactly where the resolution problems might lie, but it should be easy
to solve.

Stata creates and stores art as "vector graphics", about which if
you're interested, there's more below. Mostly you just need to know
that vector art is "resolution-independent," meaning that it stores
its information precisely, and then can be "downsampled" to any given
resolution you need. If you have a Stata .gph file already, then you
have the base graph you need to work from.

I'm not sure what happens when you copy from Stata into Word. I'll try
it quick....it looks like it copies it as vector art. So it looks like
the problem comes when you're creating a PDF from Word. My guess would
be that they want you to send your images separately. Is that true? If
so, this is easy enough to solve.

Most places that want high-resolution images (printers, for example),
will gladly accept vector images. So if you send them
".eps" (Encapsulated PostScript) files, they should be fine with that.
Stata can export to .eps, .wmf, .emf, all of which are vector,
although .eps is the industry standard. If they want a raster image
format (.bmp, .tiff, .jpg, .gif, .png) at high resolution, it's a
little more complicated, because Stata can't export these (it can
export .tiff and .png, but not at high resolution).

I just looked at the author guidelines, and the Economics of
Transition's site says "Figures should be originated in a drawing
package and saved as an EPS file."

So yeah. Open it in Stata, save it as a .eps file, and you should be
good.

Hope that helps!


Finally, a quick primer which you can read or ignore, but will help
with all this sort of stuff in the future.
---------------------------
Two ways to store visual information: vector and raster.

Raster art stores pixels as information. The most basic (and largest
by size) version is bitmap, where it essentially starts at the top-
left pixel, stores the color, and then moves to the right, stores the
color, so on and so forth, line by line, to the bottom. More
complicated versions will use image compression, sometimes with loss,
by using math. You can use principal components analysis (PCA) to
determine the main axes of variation across pixels, and drop minor
axes as needed. Additionally, if large regions are the same color, or
close, you can store a block as though it were a single large pixel.
The methods vary, but the upshot is: raster stores pixels. File size
is determined by the number of pixels and the depth of color
information (possible colors aka bits stored per pixel). The image
size determines the printed size. DPI is pixel density, so it's
roughly proportional to file size/image size.

Raster art, when you look closely, looks "jaggy" because of
"aliasing". This is because when you take a continuous mathematical
object and discretize it, you lose information. Consequently, any of a
number of original objects could have been the original source of the
new, downsampled object. Different underlying distributions fit the
same sample. In the most extreme case, imagine you reduce an image of
a person to a single, 16-color pixel. They become indistinguishable
from millions or billions of other objects (essentially, it's reducing
them to a single discrete variable = {0,...,15}). When you pixelate
letters, or circles, for example, they become indistinguishable, in
particular, from pixelated approximations of circles. You don't see
the original, intended object, but an "alias". This is a feature/bug
of raster images. It's a result of the downsampling process.

All screens display pixels. The way they avoid the jagged look is by
using a process called "anti-aliasing." Instead of pure downsampling,
the algorithm selectively removes information that will cause problems
before downsampling. In addition, it interpolates intermediate values
on high-contrast boundaries to smooth the transition.

Vector art stores the elements of a drawing as information. So it
starts with a basic image size, then throws down some cartesian
coordinates. If there's a line, or a point, or a rectangle, it stores
endpoints, line weights, line color, fill color, etc. There are no
"pixels" in vector art, just the mathematical representation of the
artwork itself. Different programs can allow more or less complex
amounts and types of art to be included, but again, the upshot is:
vector stores elements.
File size is determined by the number of points and lines and colors
and fonts used.

There is no "downsampling" in vector art, unless you are actually
removing points or lines or axes of curvature (which you can do in
Illustrator, for example, by using Reduce Points, or Smooth Lines).
Consequently, you get no such artifacts or sampling problems.
(Although, reader beware, when something like Stata or R creates
vector graphics, if you open them in Illustrator, they are MESSY.
Often, there will be multiple overprinting points. There are generally
no layers or groups, and so it's just a lot of sticks and circles and
x's all just floating in the ether.)

Examples of programs/devices that would create the different
varieties:

Raster
----------
Paintbrush
Paint Shop Pro
Digital Cameras
Scanners

Vector
---------
Adobe Illustrator
Microsoft Office Drawing
Stata

[Gustavo Javier Canavire Bacarreza]

Just save in EPS format  in Stata Denvil.

G

[Mike Blakeney]

Jason,

This may be the most interesting thing I've ever learned from an economist. Thanks!

Mike B

--
Mike Blakeney, FSA
Giraffe Whisperer

[Denvil]

Thanks Jason. I went with the eps extension and it worked well.

I had to learn the hard way just how important graphic design is for economists!!  Writing a paper is hard .. getting it published is even harder ... i was expecting the actual publication process (after acceptance) to be a cake walk, but no ... not with the ET!!!

Everyone on this forum should make time to learn about creating high resolution graphs.

Cheers!
Denvil.