Hi all, I did a quick forum and google search on this topic, but was unable to find anything about it. Is it at all possible to label individual points on a scatterplot (i.e., with the row name?)? What about clusters (e.g., using the cluster labels produced via K-medoids)? As far as I can tell, there doesn't seem to be any options for this in either the default scatterplot routine or the J-free charts scatterplotter... I know how to use the various "property" modules to change the colors and symbols, but am wondering if something similar exists (or can be built) to offer point labels too? This would GREATLY facilitate my workflow, which relies on iteritive manual manipulations of input variables, informed by identification of clusters on MDS scatterplots. Seeing where specific cases plot on these output charts would greatly facilitate this proccess. Right now, I've got to export the table with the cluster and MDS results, and then use a separate piece of software to make the plots, which isn't ideal...
Well, I've *sort of* answered my own question, as I've realized I can accomplish this using the R integration nodes. Making a custom R view with labels does work well for my purposes, but IMO having a native KNIME "point labels" node would make a lot of sense. It would also simplify my workflow, and would be attractive to those who use KNIME because they don't know (or don't want to learn) R. Labels would not have to be particularly nice; just some "quick and dirty" text labels would go miles to helping interpret spatial patterning on scatter plots. "Publication quality" plots could then still be made with an R view node.... Just my 2 cents...
Point well taken, I personally use R for most of my plotting applications in KNIME but I understand that this is not for everyone. Regarding a way to do some Quick and Dirty (c) plotting in KNIME, have you looked at using the Color, Size, and Shape managers to provide a bit more information about the data in your plots? I find this is often good enough for making quick evaluations on my data.
Labeling individual points would be a very useful feature of KNIME plotting nodes!
Thank you for your reply. I do indeed use these tools to make more sense of the plots I make with the native KNIME plotting nodes. They are very handy indeed, and do help. I also use interactive highliting between the table and plot views, which is also very very handy. However, there are some times when I just want to have all the labels on the plot so I can get a gestalt view of where all the specific cases plot. This wouldn't have to be particularly elegant looking, as I mainly use the native KNIME plotting nodes to monitor the set of results from the current workflow configuration. I, too, use R to make all publication quality graphs when the analysis is finalized...
Using R has helped me make a solution that works, but I do know that a lot of the user-base (and potential future user-base) of KNIME has chosen it because of it's intuitive graphical programing. These folks likely use KNIME instead of R because they don't want to have to write code to get things to work. This is the main reason I'm doing most of my analyses in KNIME instead of R, especially when doing dynamic analyses that require a lot of interactive tweaking of variables...
These are all just my 2 cents, but IMO I think a lot of other folks would benefit from having a plot labeling node...
Has there been any changes to labeling scatter plots point from the row name in your data table? I am currently looking into this and wondered if there has been any changes.
Hi @michaellis, unfortunately I gave up on plots in KNIME some time ago, and have now transitioned fully into Python and scikitlearn for these analyses. Making plots with Seaborn in Python are beautiful, intuitive, and easy.