Excel experts needed | Bitcointalksearch.org

organofcorti

donator

Activity: 2058

Merit: 1007

Poor impulse control.

Quote from: bb113 on March 02, 2012, 06:00:26 PM

So you have a script that creates this automatically from the data in R?

I was annoyed I hadn't been able to come up with an answer so I read up about Grid and ggplot2 and came up with this last night. It might need some tweaks depending on your data, plus the axis flip only works one way which is why I put the vertical graphs on the right instead of the left side.

But, yes, using Grid to place your graphs is doable, and if you want different triangle sizes or different shapes or different width/height graphs you can change those params too.

Quote

And I should be able to warp the shape?

I'm assuming you know what you're doing with svgs because I have no clue. If svgs are warpable, then I guess you can warp it. PM me with an email address and I'll email you the svg I did when I get home tonight and you can test it.

Quote

I think mine looks alot prettier but your way may be more functional.

Nuh-uh. I'm way prett- er I mean mine's way prettier.

Anything you don't like the look of is probably easy to fix.

bb113

hero member

Activity: 728

Merit: 500

So you have a script that creates this automatically from the data in R? And I should be able to warp the shape? I think mine looks alot prettier but your way may be more functional.

organofcorti

donator

Activity: 2058

Merit: 1007

Poor impulse control.

Quote from: bb113 on February 25, 2012, 01:25:55 AM

Looks like this:

how bout this?

the vertical graph limits are slightly out, but i don't have time to fix - bitclockers fires to put out.

i can email you the svg if you like. Alpha doesn't work on the svg though.

bb113

hero member

Activity: 728

Merit: 500

I have messed around with blender a bit in the past, are you saying I could make a stack of images (about 30) and plot my data points on each accordingly, then the software can create a 3-D model of this? If so the Dicom step would be unnecessary.

marked

full member

Activity: 168

Merit: 100

Quote from: bb113 on February 25, 2012, 03:02:03 PM

3-d with resolution of 200 um along the z-axis (the data is from 200 micron thick coronal sections). I have no idea about dicom viewer.

I should note that this is aggregate data from multiple brains, so it would need to be overlayed over a generic stack of images. I was going to just do 2-d overlays on a coronal section and saggital section, but a stack of coronals would be cool if it allowed me to get some kind of perspective snapshot.

DICOM is a standardised set of imaging formats for medical imaging in 2d, 3d and 4d (and apparently 5d). A DICOM viewer can pull images from PACS, local storage, etc.

Perhaps a quick scan of the screenshots and video at http://www.osirix-viewer.com/Snapshots.html (OSX dicom viewer) as to what you could be looking at, and then delve into the material listed from http://en.wikipedia.org/wiki/DICOM

My understanding is that you can build your representation in a format that can be applied over another image or combined then imported and viewed. I'm not entirely certain as to what the initial representation formats can be, if it is something simple and can be programatically built in a 3d toolset such as Blender3D. Once imported it can be treated as any 3d object, and if needed can be seen in time (an example of this would be a false colour beating heart.)

As I don't have a medical background, and have only really looked into this as a "what 3d tools can be used in healthcare" on an android phone, I don't know whether I can contribute much more. I primarily suggested it as you might be able to go back to original source for the images as they would be the original export or at least first intermediate format, particularly as you said that you were warping the data image somehow.

marked

bb113

hero member

Activity: 728

Merit: 500

3-d with resolution of 200 um along the z-axis (the data is from 200 micron thick coronal sections). I have no idea about dicom viewer.

I should note that this is aggregate data from multiple brains, so it would need to be overlayed over a generic stack of images. I was going to just do 2-d overlays on a coronal section and saggital section, but a stack of coronals would be cool if it allowed me to get some kind of perspective snapshot.

marked

full member

Activity: 168

Merit: 100

Quote

The main thing is that wanted to export with vectors intact so I could layer it over a picture of a brain slice then warp the chart to fit the curvature of the brain. A screenshot is ok for any other use. Actually I do want to make it 3-d (include z-axis info), I don't think this will be possible in excel though.

Is there a way you can export this information into a 3d representation that can then be imported into a dicom viewer? is the brainslice 2d or 3d (and what thickness?).

marked

bb113

hero member

Activity: 728

Merit: 500

The data isn't that big a secret im just being paranoid. I don't think a contour plot would work, the scatter plot is actually a 2-d map of neurons, the size of the triangles is the size of the neurons and the colors are different categories of neurons. The lines show how the size of the neurons differs according to location. The bar charts just show the sample size at different locations so I know how much to trust the data at different levels of x and y position. It is relatively intuitive if you know what you are looking at.

The main thing is that wanted to export with vectors intact so I could layer it over a picture of a brain slice then warp the chart to fit the curvature of the brain. A screenshot is ok for any other use. Actually I do want to make it 3-d (include z-axis info), I don't think this will be possible in excel though.

organofcorti

donator

Activity: 2058

Merit: 1007

Poor impulse control.

Is it for you or other people?

Usually when I present data of several categories and three dimensions I'd use several colour maps or contour plots to show trends. I find it hard to follow your plots even though you probably understand them quite well.

So my quick answer is I know of no easy way to do what you want in R. There probably is some plot method that's flexible enough, but I don't know it.

If you don't mind the data being presented in a different way that still gets your point across (normalised if you don't want me know know what the number you're sending me actually are) I'll be happy to have a look.

Also, if you want to make the best impact possible with your data you should look at some of the books and websites that deal with how data presentation affects your audiences ability to grasp what they are looking at.

Good luck with it.

bb113

hero member

Activity: 728

Merit: 500

Like I said, a complex chart...

bb113

hero member

Activity: 728

Merit: 500

I dunno I made it as I explored my data. Its a scatter-"bubble" plot (x and y variables plotted normally, then z as size of the data point, and color as a category). Then I plotted the values of the z variable as a functions of the x and y variables, placed those charts accordingly, next took a simple moving average to smooth them (the black lines) to look for overall trends. Then I wanted to know how much to trust the smoothed line at different points and how much each category contributed so I put a stacked count of that "underneath".

organofcorti

donator

Activity: 2058

Merit: 1007

Poor impulse control.

What is that kind of chart called?

bb113

hero member

Activity: 728

Merit: 500

Looks like this:

How well does R work with custom fonts (bolding, italics, symbols, etc)

organofcorti

donator

Activity: 2058

Merit: 1007

Poor impulse control.

Quote from: bb113 on February 25, 2012, 01:12:07 AM

Yea, screenshot is what I did earlier. I have seen the graphs out of R, I don't think it would be as easy to get the aesthetic I want. Can I export combo graphs from R to illustrator preserving vectors and text?

http://cran.r-project.org/web/views/Graphics.html

"RSvgDevice is an SVG device driver and interfaces well with with vector drawing programs, or R web development package"

If you can post something along the lines of what you want I can tell you if it's easy to do it in R or not.

bb113

hero member

Activity: 728

Merit: 500

Yea, screenshot is what I did earlier. I have seen the graphs out of R, I don't think it would be as easy to get the aesthetic I want. Can I export combo graphs from R to illustrator preserving vectors and text?

deslok

sr. member

Activity: 462

Merit: 250

It's all about the game, and how you play it

you can create charts with multiple lines assuming you have one common axis otherwise the options to combine seperate charts using seperate datasets are limited, that being said combining screenshots is simple lining things up should be easy enough in the right program but you really are better off with a seperate graphing program for the level of complexity you might be better off with seperate software

ColdHardMetal

hero member

Activity: 700

Merit: 500

Just do a screenshot of excel if they are set up and looking the way you want, and all you need is a jpeg.

organofcorti

donator

Activity: 2058

Merit: 1007

Poor impulse control.

It doesn't need to be excel - that's limited. R is much better at graph plotting. What do you need?

bb113

hero member

Activity: 728

Merit: 500

So I made a complex chart (don't want to link it because it contains sensitive info) that consists of multiple charts overlayed on each other. Is there a way to export this combination of charts as a pdf or jpeg, or whatever (better if it preserves vectors and text). I could export each chart to illustrator individually, but there are 7 different charts involved and it is a pain to align them all correctly. Is there a way to group them all (even if some are hidden behind others) so that they print and copy together?

Topic: Excel experts needed (Read 1555 times)