Announcement

Collapse
No announcement yet.
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • stripplot updated on SSC

    Thanks to Kit Baum as ever, stripplot has been updated on SSC. The code has not been updated since 2014, but the help file is bigger and shows a lot of rewriting. Some of the size reflects a personal habit of letting a help file be. in Frederick Mosteller's terms, the zeroth draft of a paper.

    Stata 8.2 is needed. That's not demanding in 2016.

    To install, use

    Code:
    ssc install stripplot
    and to update use ssc or adoupdate.

    In this case, the project, or perhaps paper, has been evolving since 1999. With good fortune, I may get to writing some of it up in the Stata Journal in the next couple of years. stripplot started out in life as onewplot, that name itself reflecting the filename.ext or 8.3 rule for filenames that bit under MS/DOS and so limited Stata filenames. (I regularly come across young people who have no idea what MS/DOS means, but that's fine. They don't need to punch 80 column cards either.)

    onewplot was intended to do some things that the then graph, oneway could not do, or not do well. To see what graph, oneway does you can call up

    Code:
    sysuse auto
    sort rep78
    graph7 mpg, by(rep78) oneway
    But after that a big twist in the road led towards the present, and the best short summary now is that stripplot is, functionally, a superset of the official command dotplot. (It never started with a clone of that code, but it is now aimed to do what dotplot does, and more.)

    The plain or vanilla uses of stripplot are fairly mundane, but its best uses lie in combining options, as when miniature quantile plots have boxes for median and quartiles and reference lines for means superimposed. (On the latter, searching here for "quantile-box" will yield examples.)

    The preamble to the help has been restructured, and the examples can be run in sequence using Stata's own datasets.

    The list of references is elephantine by some standards, and shows up a personal pleasure in discovering examples of, and variation on, strip plots under many other names in literature old and new, and in odd corners too. There is plenty of scope for more references.


    Philip Ender
    William Dupont
    Kit Baum
    Maarten Buis
    Ronán Conroy
    Marc Kaulisch
    David Airey
    Oliver Jones
    Fredrik Norström
    Marcello Pagano
    Dionyssios Mintzopoulos
    Judith Abrams
    Vince Wiggins
    Frank Harrell
    Alona Armstrong

    variously provided bug reports, specific suggestions, pertinent references and encouraging noises, which I am more than happy to acknowledge once more.




  • #2
    For those like me who might still know what MS-DOS means, but have no idea what a stripplot is, here's just one of the examples in the help file, showing temperatures for various cities per census region and season. If I'm not mistaken, the red dot shows the mean, the blue extensions are the 95 (?)% confidence intervals.

    Click image for larger version

Name:	stripplot_example.png
Views:	1
Size:	100.9 KB
ID:	1362877

    Comment


    • #3
      Jesse is correct on all points.

      Comment


      • #4
        Dear Nick

        Newbie question. :-) I'm baffled with stripplot on my mac! I use Stata on a windows desktop (mostly), but also on my macbook pro (on the road). They are both MP/14.2. I am able to create beautiful stripplots in my windows version, but for some reason on my mac, I cannot see the box graph unless it is filled in (bfcolor specified). It also refuses to draw the lines for IQR or mark the median. I have been using the same dataset in both cases, but to be sure, I also tested it with the auto dataset and I get the same frustrating result. For example, if I type:

        stripplot mpg, over(foreign) cumul cumprob box centre vertical

        I get

        I must be missing something but could not figure out. Let me know if you have any tips/advice.

        Thank you,
        Nazli
        Attached Files
        Last edited by Nazli Uludere Aragon; 17 Dec 2016, 22:28. Reason: I have an insight: I discovered in graph edit that for the box components the default outlline widths were set to "vvv thin". I wonder why? And how can I change the defaults to "thin&qu

        Comment


        • #5
          Sorry, but I don't have a Macintosh to check that. I don't get it under Windows. There is no setting "vvv thin" within stripplot. That's not even a documented line width setting: http://www.stata.com/help.cgi?linewidthstyle

          I'd check what you get with e.g.

          Code:
          set scheme s1color 

          Comment


          • #6
            Updated again, thanks again to Kit Baum!

            Update with

            Code:
            ssc inst stripplot, replace
            or adoupdate.

            The update concerns (1) improved treatment of reference lines, now available in conjunction with by() (2) further updates to the help.

            The problem reported in #4 was also reported in http://www.statalist.org/forums/foru...ata-14-for-mac As discussed there, I don't believe that the problem is a bug in my code. The problem arises only on Macintosh, and there is no OS-dependent code in my program.

            Comment


            • #7
              Thanks to Kit Baum again, a new version (2.7.0 2 March 2017) is now available from SSC. Use ssc or adoupdate to install if interested.

              Full story within http://www.statalist.org/forums/foru...ata-14-for-mac

              Comment


              • #8
                Thank you for this. I was wondering if there is a way to circumvent the limitations in using by and over with multiple variables. I appreciate stripplot's the ability to show individual data-points in a boxplot and I would like to do so whilst comparing categories (for example different group of patients) AND variables (for example different biomarkers).
                Thank you so much

                Comment


                • #9
                  You need to reshape your data first. Only way to do it, I think, short of writing your own command.

                  Comment

                  Working...
                  X