Apache OpenOffice (AOO) Bugzilla – Issue 5289
TrendLine/graphing functionality in OpenOffice
Last modified: 2013-02-24 21:21:48 UTC
I believe that the trendline/regression curve capabilities in OpenOffice at the moment is very limited. I think that this area needs major improvements. Firstly at the moment, it is impossible to extend a trendline beyond the current dataset to extrapolate results outside the given range of input data. This function would be very useful at least in scientific writeups and most probably in many other areas as well. Secondly, the current method of creating a trendline is very limited - by simply selecting which type of trendline to use and it simply appearing. It would be much more functional if a "function" existed to "add a trendline" which then became a seperate object linked back to the data set from which it was created. This would allow two things - firstly multiple trendlines could be created - secondly trendlines could be "deleted" using the delete key (at the moment this is impossible, one has to go into the dataset options and turn off the regressions curve to get rid of it, this is not an intuative way of doing things) - and thirdly hopefully a feature could be added so that the equation of a trendlsetine could be displayed, also a feature which would be very useful. Thirdly in the area of uncertainty display on the graphs. When selecting a percentage error on a dataset, you cannot specify any more accuracy that integer percentage errors. Even worse, for constant error margins, you cannot specify anything more accurate than one decimal place - and for some datasets this is just not sufficient as the range of data is very small and the error margin MUST also be the proper value, not some number rounded to one decimal place when perhaps the entire area of the graph fits withing one deciamal place. Well, I hope someone will be able to take this on, as on the whole I am very impressed with OpenOffice, it is just these small things which at the moment prevent it from competing successefully with the giants of the industry - by this I mean that I am FORCED against my will to boot up into windows and use Microsoft Office for graphing projects when I would prefer to be using OpenOffice.
Hi Falko, 1 4 u Frank
I'd just like to add my support for this issue, as well as outline my hopes and dreams for OpenOffice's trendline support. I use spreadsheets mainly for data analysis (just finished BS's in chemistry and biology, working on masters), so basic statistical functions are very important to me. I think that the following features are kind of a bare minimum of functionality: 1. Trendlines be easy to add (they are, but an option on the context menu would be nice), and have their own properties dialog (it is much more intuitive). The properties dialog would include the trendline type. As the above poster suggested, it would be great if they were their own object. 2. Trendlines have the ability to display the line function and R^2 value as an object in the graph (shouldn't be hard, as the chart obviously calculates them...just need to display them). This is *extremely* important for any kind of meaningful data analysis. It would also be great if the user could adjust the properties of the text box displaying the equation and R^2 value, so that the user could change the number of digits displayed, scientific/general notation, etc. 3. Trendlines have the ability to extrapolate ahead/behind the data an arbitrary distance. As kind of a future wishlist, a few other things would be nice to have, and would give OpenOffice some pretty heavy-duty stats abilities. 1. Allow trendlines to consider only a *part* of the data set, perhaps by setting the data range to consider in the trendline properties dialog. This allows one to use two or three different trendlines on a single data set, which is very useful in some situations. For example, if you have a data set that starts out constant, then suddenly moves to a different value and stays constant there, you can generate trendlines for the low constant, the up-slope, and the new constant. This is extremely useful in thermodynamics, for example. I've only seen this functionality in fairly expensive, propriatary stats packages. 2. When the cursor is positioned over a data point, have the data point coordinates "pop-up" in a help-menu type manner, which makes it much easier to figure out what point you are looking at. Anyway, I hope that someone has time to implement some of the basic trendline stuff. It is currently completely impossible to write a lab report for an undergraduate science class using OpenOffice, as every conceivable lab would involve examining trendlines of data sets. I detest making my graphs on another machine and printing them out seperately!
*** Issue 5085 has been marked as a duplicate of this issue. ***
Hi, Related to this issue, besides the points already mentioned, a "Polynomial Regression" would be a very useful addition to the OpenOffice Chart.
/me especially wants to extend the trendlines and using only part of the data...
setting target-milestone... adjusting priority.
From what I understand of this issue I consider this feature a good-to-have. Therefore re-assigned to Bettina for feature commision.
Changed to started
*** Issue 20819 has been marked as a duplicate of this issue. ***
Hello Björn, please give approval for this evaluated OO.o 2.0 flagged issue. If you confirm with the target OO.o 2.0, then please keep it on your owner (or the owner of the concerning developer) for implementation. In case you want this issue for 'OOo Later', then please reset the target milestone. If you decline the issue finally, please set the resolution to 'Wontfix' (but do not close). In case of 'OOo Later' or 'Wontfix' please reset it on Bettina's owner. Thank you.
In the new chart implementation, trend-lines are objects. You can (technically) add more than one trend-line, you can obtain a string representing the formula and you can get the r^2 value. However (!) the UI still is the same, i.e., you can still have only one trendline, and the formulas will no longer be displayed in the future (currently they are as a test). The issue about the precision of the constant error values is in my opinion a bug. I would suggest to change the metric field (which contains an integer that is interpreted as tenths) to an edit field which can hold different formats like the axis-scaling fields (min, max, etc.). I suppose, this bug will not be fixed entirely for OOo 2.0, but maybe in parts.
Forgot to set state to 'Accepted' aka 'STARTED'.
according to the announcement on releases (http://www.openoffice.org/servlets/ReadMsg?list=releases&msgNo=7503) this issue will be re-targeted to OOo Later.
setting keywords, subcomponent according to new RFE-process.
Another important feature, like issue 366, re-targeted to OOo Later...I don't understand. This feature is essential to scientific workers. Please, reconsider it. Regards, ap78
These guys JUST DON'T GET IT! No matter how many bugs/issues get filled about error/statistical analysis/trending they continue to view these as "enhancements". Get your heads out of your ***** and recognize that the scientific/financial community is a sizeable portion of spreadsheet users and neither of them even look at OOCalc because these features are missing. I get it, you are programmers and don't need these things but you are providing a product for your community and we are SCREAMING ABOUT THE LACK OF THIS FEATURE. I have generously given funding to other OSS projects but REFUSE to give anything to OO because of their ability to turn a deaf ear to this part of the community. Look at issues 23909, 22740, 25277, 28458, 40578 and 366. I for one would rather have these features than any performance boost or prettying up of the interface.
Maybe I wouldn't use such strong language, but enmane is right. The lack of decent scientific/finance capabilities makes OOo completely unsuitable for any kind of scientific work. As much as I'd love to roll our labs over to Linux/OOo, we just have to have Excel around to do any kind of data analysis at all. From what I've seen of the 2.0 candidate, nothing will be changing for years, so we're still forced to stick with MS Office. I would gladly pay for a usable replacement, but there just isn't one, and OOo isn't moving in that direction, despite the vocal requests of the scientific and financial community. No, I can't 'code those features myself or shut up', I'm a geneticist, not a programmer. The best I can do is community support via bug reports/new user assistance and financial support.
Believe me, we have not chosen to ignore your concerns. As you might know (it is even mentioned in this issue) we have re-written the entire chart module with the intention to enable it with a lot of important features, including better regression curve capabilities. The new implementation exists and already contains better regression curve interfaces, but it is far from being complete. Our intention was to bring the new chart into OOo 2.0, but changed priorities and deadlines lead to postponing the new chart. As we had the focus on the new implementation, logically the old chart (that comes with OOo 2.0) will not have changed much. We hope that we will be able to continue the new chart implementation soon.
Adding myself to CC since this is also of great importance to me.
Hi folks, this issue gots much to complex to work with. There are a lot of good feature requests in there but so many that this issue will likely never be finished. Also nobody knows for which feature excatly the other one has given his vote. I reopened some of the duplicates already because they do concentrate on a single problem each. That are Issues one can work with: -Issue 5085: extrapolation of regression curves. -Issue 7998: display of regression curve parameters. -Issue 20819: add polynomial regression type. -Issue 8877: allow constant error bar values smaller than 0.1. Please help with creating separate issues for each feature or defect. Thanks.
Adding myself to CC.
*** Issue 49600 has been marked as a duplicate of this issue. ***
Yawn... Say fellas, it's been 3 yrs. How much longer? Just curious and a little more than impatient at this point. Anything we need to do to make this a political MUST DO? How much money would it take to get you guys to jump on this?
I share your frustrations! I haven't posted on this issue since I first created it 4 years ago. Back then I was in high school. Now i'm nearing my final year of University. From the comments over the years it seems as though some of the backend work has been done - but it would be nice for this work to bear some fruit for the user some time soon! David
The chart reimplementation is still ongoing and not finished yet. You can download a developer snapshot at the chart website if you like: http://graphics.openoffice.org/chart/chart.html. We are preparing a new download currently and will announce it on the openoffice.graphics.dev mailinglist.
I don't find this issue very helpful as it contains too many sub-problems. ->IHA: As discussed to you.
I tested the new chart implementation, but I was surprised not to see any check box or button to show the equation of the regression curve. I know the chart module is still under development, but this feature is widely used in Excel and I think it should be present. Will it be implemented? Regards, Alberto
Schools is the place where people is meeting new software, but OOo is a lot less usefull than Excel in math-education, as long as you have to use LINREGR() to see regression coefficients. In Excel they appears in the chart after a few clicks. Please make something more easy to use than Excel, or my pupils will not consider using OOo at all. Also a better interface for changing line width, point size and colors would be a major step forward. These things seems trivial, but matters for teenagers. Yours Mads _*_
*** Issue 71475 has been marked as a duplicate of this issue. ***
I definitely would like to be able to adjust the width and color of trendlines. I occasionally have to compare regressions, so allowing multiple trendlines for one function would be nice, as well as being able to see their equations and R^2 values. Polynomial regression would be nice. Harder to do, but very useful, would be logistic regression.
*** Issue 76161 has been marked as a duplicate of this issue. ***
I would say that the most important part of this feature is the display of the equation ON the graph. It's nice that it's finally on the screen, but it needs to be in the picture. Also, when dealing with trendlines, pressing 'delete' should remove them, rather than cause an error. Overall, the change in 2.3 was a step in the right direction, but it didn't get all the way there.
With the coming OOo 2.4 equations for regression curves can be displayed (issue 7998), trend lines can be added from the context menu and they can be delete via delete key. The precision problem for the error bar values was already addressed in OOo 2.3 I think. If you want to try the current state you are welcome to download a developer snapshot here: http://download.openoffice.org/680/index.html?intcmp=1235 Going further to version OOo 3.0 there is already more work done. With issue 44768 inserting a trend line from the context menu immediately shows a properties dialog where you can choose the type. Have a look at the specification for the current state http://specs.openoffice.org/chart/Trendlines.odt . The fix is not introduced in the master trunk right now - look for the integration of CWS chart19. There are still several known open issues in the area of trend lines and error bars. Please help identifying the remaining hot spots and move your votes to the concrete issues: Issue 5085 extrapolate regression lines (extend trendlines) (forward/backward forecast) Issue 20819 add polynomial regression type Issue 34093 Force linear regression curve to go through zero point (origin) Issue 64813 Allow more than one regression line for one series Issue 66819 Regression curves: new type "moving average" Issue 366 Select 'Range' for error bars (work is in progress) Issue 5787 easy creation of charts with errorbars Issue 42946 Error Bars for X values Issue 64816 Mean value for X and center of gravity Issue 66788 Error bars: Standard deviation based on sample Issue 69664 Poisson error bars Thanks for your input! I close this issues as duplicate now. *** This issue has been marked as a duplicate of 7998 ***
closing