| View previous topic :: View next topic |
| Author |
Message |
ousouji Newbie

Joined: 03 Nov 2004 Posts: 1
|
Posted: Wed Nov 03, 2004 6:17 am Post subject: Specifying cells for error bars |
|
|
I'm not sure if there's already a thread about this. Howdy. I'm new and I use slack 10.
Anyway, I'm often presented with the issue of multiple data sets for y values, and a plot of their averages. Each individual point on the plot has a unique standard deviation based on these (original) multiple data sets. What I'd like to be able to do is specify a set of cells that contain the standard deviations, so that the error bars are specific to each data point. The stats tab includes the ability to add error bars, but these seem to only be customizable to a fixed variance for all the plot points. Otherwise it calculates standard deviations based on an average of all the plot points, which is counterintuitive to what I usually want to do.
I'm not sure if I can do this or not. I know this is doable in excel so half the time I have to resort to some windows machine somewhere, which I really don't like doing because it means I have to leave my comfy swivel chair.
Again sorry if there's already a thread about this and if someone could point me to it, that would rock.
Thanks.
((Oh and as an irrelevant plus, the math writer is sooooo awesome. A billion times faster than the clunky point-click-point-click of msword's formula editor.)) |
|
| Back to top |
|
 |
gerwulf Newbie

Joined: 01 Jun 2005 Posts: 3
|
Posted: Wed Jun 01, 2005 8:30 am Post subject: |
|
|
That's exactly what I am looking for, too. My girlfriend is using OO for writing the thesis of her doctorate, but she has to switch to MS-Excel for this exact functionality.
Since I'm already able to set a fixed or percent value for the standard deviation (using 1.9.104beta), I think it would be fairly easy to implement the option of specifying a certain cell thats holding the intended value.
On the other hand, I'm only a very noobie-coder of sorts, so my judgement may be misleading me.
Anyways, it would be really great to get this feature into what's going to be OO 2.0
Gerwulf
Hail to all those great people that make things like OO possible! Praise to you!  |
|
| Back to top |
|
 |
joerg04 OOo Advocate


Joined: 14 May 2004 Posts: 326 Location: Germany
|
|
| Back to top |
|
 |
David Super User


Joined: 24 Oct 2003 Posts: 5668 Location: Canada
|
Posted: Wed Jun 01, 2005 1:21 pm Post subject: Re: Specifying cells for error bars |
|
|
| ousouji wrote: | What I'd like to be able to do is specify a set of cells that contain the standard deviations, so that the error bars are specific to each data point.
...
|
Do you mean you'd like to see "relative" error bars for a set of points? I know of a freeware program if that is what you need. You can set the same % range for each point-bar.
David. |
|
| Back to top |
|
 |
9point9 Moderator

Joined: 31 Aug 2004 Posts: 3875 Location: UK
|
Posted: Wed Jun 01, 2005 2:07 pm Post subject: Re: Specifying cells for error bars |
|
|
[quote="David"] | ousouji wrote: | | Do you mean you'd like to see "relative" error bars for a set of points? I know of a freeware program if that is what you need. You can set the same % range for each point-bar. |
I don't think that is what is meant. I think ite means using a calculated standard uncertainty for each point which in many cases will vary between points. As a physicist I use gnuplot to do this sort of stuff, I've written a macro for OOo Calc 1.9.x that allows data to be exported from a spreadsheet to gnuplot, resulting in a PNG image of the graph. Means that I don't have to mess around with gnuplot scripts and is far quicker than plotting a graph and setting all of the options under under spreadsheet. _________________ Arch Linux
OOo 3.2.0
OOoSVN, change control for OOo documents:
http://sourceforge.net/projects/ooosvn/ |
|
| Back to top |
|
 |
David Super User


Joined: 24 Oct 2003 Posts: 5668 Location: Canada
|
Posted: Wed Jun 01, 2005 3:56 pm Post subject: Re: Specifying cells for error bars |
|
|
[quote="9point9"] | David wrote: | | ousouji wrote: | | Do you mean you'd like to see "relative" error bars for a set of points? I know of a freeware program if that is what you need. You can set the same % range for each point-bar. |
I don't think that is what is meant. I think ite means using a calculated standard uncertainty for each point which in many cases will vary between points. |
Forgive me, but I'm getting lost between "standard" and "vary". No matter; if you and he understand each other that will do.
David. |
|
| Back to top |
|
 |
9point9 Moderator

Joined: 31 Aug 2004 Posts: 3875 Location: UK
|
Posted: Wed Jun 01, 2005 4:18 pm Post subject: Re: Specifying cells for error bars |
|
|
| David wrote: | Forgive me, but I'm getting lost between "standard" and "vary". No matter; if you and he understand each other that will do. |
It's statistics, I don't blame you for getting lost. Basically if you have a set of data points with different y values but all one x value then you can calculate the mean and the uncertainty in that is equal to sigma/sqrt n. This is what you plot as a data point with y-error bars. You can then use something like the non-linear least sqares Marquardt Levenberg Algorithm to fit a straight line to it with uncertainties in dy/dx and y intercept. Unfortunately OOo Calc doesn't do this which is a shame as it is in many other ways the best office suite for science. _________________ Arch Linux
OOo 3.2.0
OOoSVN, change control for OOo documents:
http://sourceforge.net/projects/ooosvn/ |
|
| Back to top |
|
 |
gerwulf Newbie

Joined: 01 Jun 2005 Posts: 3
|
Posted: Thu Jun 02, 2005 1:22 am Post subject: |
|
|
Thanks everyone for your replies.
I am and have been aware of R/gnuplot/gnumeric and the like, and personally have used/would use them again, even if that meant some additional hassles with installing/getting everything to work as it should.
On the other hand, most people (like my girlfriend) are quite unwilling to learn handling yet another software, and don't even mention finding/installing/configuring it.
For those people (the vast majority of today's PC users) it's often just a simple decision:
Use OO plus R/gnuplot/gnumeric, including install/configuration/learning to use them, or just stay/switch back to MS-Excel, which can do just what they want without any changes.
Correct, Excel may not be able to do any other things better than OO, but that would still be sufficient to drive 'normal' users back to Excel if he/she is missing this exact functionality.
All I'm saying is that if OO wants to get really widespread, I think it has no choice but to be able to do everything Excel can, and then some. Agreed, not everything Excel can do is worth imitating, but useful options (like this one) really are, IMHO.
Btw: is there any central or better suited place where I could bring this 'feature request' to the attention of the coders involved in Calc (without embarassing them), or is this already the best place to do so?
And 9point9: is there a place where I can find the macro you mentioned?  |
|
| Back to top |
|
 |
ftack Moderator


Joined: 27 Jan 2003 Posts: 3102 Location: Belgium
|
Posted: Thu Jun 02, 2005 2:42 am Post subject: |
|
|
I agree fully with your concerns. It is also known and recognised that the charting in OOo is not up to current standards. There have been serious efforts to have an entirely rewritten charting engine ready for OOo 2, but they have not been able to make it. In the future, we surely will see a significantly improved charting, but obviousy, that does not help todays needs.
To actively contributue in filling "issues" (bugs, enhancement requests, etc), you just need to register on the official OOo web page, www.openoffice.org. First read a bit the recommendations before starting to submit issues. The issue you mention here surely has been filed. Once you are a registered user, you can "vote" for it to help saying that this issue is particularly important to you. |
|
| Back to top |
|
 |
gerwulf Newbie

Joined: 01 Jun 2005 Posts: 3
|
Posted: Fri Jun 03, 2005 2:45 am Post subject: |
|
|
Thanks for the directions, ftack.
I registered and found that the issue 3380 describes what we are talking about here: http://qa.openoffice.org/issues/show_bug.cgi?id=3380
To everyone else here who would like this to be implemented soon: register on www.openoffice.org, use the 'My Pages' link in the top bar, then the 'Find issue' link at the left, enter '3380' in the 'Jump to issue' field.
Then, you can use the 'Vote for this issue' link further down on the page displaying issue 3380, and cast votes for it. Up to two votes are possible for a single issue in this module.
Cheers,
Gerwulf |
|
| Back to top |
|
 |
DocHusten Newbie

Joined: 14 Sep 2005 Posts: 2
|
Posted: Wed Sep 14, 2005 3:57 am Post subject: |
|
|
Hello,
I am a scientist and I am using OpenOffice/Staroffice since more then ten years. In this time I tried to convince StarDivision (the former company who produced StarCalc) to implement this function several times. Therefore I am dissapointed about this topic, because this function is essential for every data plot.
I am using an other programm (I know this is not the solution): qtiplot (http://soft.proindependent.com/qtiplot.html). This software is availbe for Linux (also rpms for Suse, Mandriva...) and Windows.
Greetings
Michael |
|
| Back to top |
|
 |
duff Power User

Joined: 14 Sep 2003 Posts: 80
|
Posted: Sun Sep 25, 2005 12:05 pm Post subject: i second the above |
|
|
use qtiplot... it does custom error bars (eg sem) easily. simply paste relevant values into a column in the qtiplot spreadsheet and select that column as the source of your error bars. oh and if you do download a binary from the homepage please donate a little to the project.
duff |
|
| Back to top |
|
 |
DocHusten Newbie

Joined: 14 Sep 2005 Posts: 2
|
Posted: Mon Sep 26, 2005 9:23 am Post subject: |
|
|
Hello,
I am using qtiplot (and I made a donation to Ian). Despite that point, I wish this essential feature for OpenOffice. qtiplot has some problems when I have more than one y value corresponding to one x value.
Grettings
Michael |
|
| Back to top |
|
 |
|