SPORTSCIENCE · sportsci.org News & Comment / In Brief

• The Best Graphing . Solving the problems with Office. • Updates: Sample size; Clinical inferences; Validity and reliability Reprint pdf · Reprint doc

The Best Graphing Software Will G Hopkins, Sport and Recreation, AUT University, Auckland, New Zealand. Email. Sportscience 15, i-iii, 2011 (sportsci.org/2011/inbrief.htm#graphs). Published June 2011. ©2011 Update Aug 2018. Bill Microsoft Gates has at 2010 has unacceptable bugs for such pro- last solved the problem (detailed below) of cessing, and my IT people are telling me I have lines turning into tram tracks when graphics to give up Excel and Powerpoint 2003 when from Excel are pasted into Powerpoint as a they install Windows 7 on my laptop soon. So metafile and ungrouped for further editing. I am looking for graphing software that will Thanks, Bill, you got there in the end! The make graphs I can modify with the latest Pow- latest version of Powerpoint in Office 2016 also erpoint. In any case, I am interested in software has a wonderful new feature for lining up ob- that does a better job than Excel, although I still jects, whereby dashed lines appear when edges have to use Excel: all my spreadsheet resources of an object you are moving line up with edges are in Excel, so anyone using the spreadsheets, of other nearby objects. Now, Bill, please rein- including me, will still want to transfer figures state Recent Folders in the Save As… windows from Excel into Powerpoint. in . Solution #1: Keep Using Office 2003 Update Feb 2018. I still use Excel to create I have Jim Martin to thank for sharing similar graphs, and I avoid the thick lines generally by problems and strongly advising me to keep not taking apart the graph after pasting into Office 2003 going. He said I could buy Office Powerpoint. Instead, I clean up the graph by 2003 for US$130, if need be, "a pretty inexpen- covering items with white boxes and adding sive solution". I agree. I will need technical new numbers, text and lines. I would use help to get 2003 and 2007/10 running together, GraphPad Prism instead of Excel, if I had it. as I did for my current setup with Windows XP. Graphs in the new free SAS Studio package can I will hire an outside geek if my IT people re- be taken apart in Powerpoint, but they require fuse to help. too much programming for the casual user. See So here's what to do when you want to use examples in the suite of materials on mixed the graphs in Excel to make graphs that look modeling in SAS Studio. good in Powerpoint presentations and that are In April this year (2011) I posted a message acceptable in publications (adapted from my in- to the Sportscience mailing list, asking for ad- brief item on preparing graphics for publica- vice on software for creating publication- tion): open the xls or xlsx spreadsheet with quality graphs. I was hoping to solve a problem Excel 2003; copy the graph to the clipboard; I had encountered with the use of the 2007 paste-special as a Windows metafile; click- Office suite for graphing, as described in an drag to change the size; right-click and un- earlier in-brief item. Here is a restatement of group twice; edit the graph elements (easier if my problem and the various solutions, based on you first delete the white background that the dozen or so responses I received and further comes through with the graph); resize again if experimentation. necessary (by ungrouping all grouped items, cutting to the clipboard, re-pasting as a meta- The Problem: Bugs in Office 2007/2010 file, re-sizing...); finally export it for publica- I have been using Excel 2003 to create tion, or copy the graph or the whole slide into graphs, which I then paste into Powerpoint Powerpoint 2007 or save the whole presentation 2003 to take apart and clean up for slideshows as a pptx for presentation. and/or manuscripts. My problem is that the The only alternative to using Powerpoint Excel+Powerpoint combo in Office 2007 and

Sportscience 15, i-iii, 2011 In Brief Page ii

2003 is to edit the Excel graphs in a vector Solutions #4 and more drawing editor before pasting into Powerpoint The open-source R package is renowned for 2007. I searched for freeware editors and tried its graphics. It's also free, incredibly powerful, Inkscape, but it doesn't take Excel graphs apart, but user-toxic when I last tried it a couple of and I don't know whether expensive high-end years ago. A Windows interface is available editors like CorelDraw and Macromedia Free- and may help newbies like me get started. Rob hand will do the job. In any case, you can re- Rein, who runs the R tutorial group, sent me an gard Powerpoint 2003 as a cheap-ware power- R graph as a Windows metafile, and it came ful drawing editor that has a great additional apart perfectly in Powerpoint 2007. advantage: the ability to resize graphics and Dan Heil was ecstatic about graphs in Statis- keep everything, including font sizes, in propor- tica. I checked with our local Statistica agent, tion, as described above. who sent me a graph that edited perfectly in By the way, one of the only advantages of Powerpoint 2007. Price: ~US$675 (I was quot- Powerpoint 2007 over 2003 is the ability to edit ed NZ$850+tax). Statistica probably represents an element of a group without ungrouping better value than SigmaPlot. Statistica even has (which in 2003 results in loss of the animation a mixed model, but it isn't good enough yet for for the group). I sometimes go into Powerpoint our purposes. I periodically ask them to im- 2007 solely for this operation, then scuttle back prove it. to 2003. It's bizarre that the Microsoft develop- Michael Brach: "I'm working with ers could have an eye for this kind of detail, yet while most of my colleagues use Windows, I'm ignore the graphing needs of the scientific always looking for software which uses open community. formats and/or is available for both operating Solution #2: GraphPad Prism systems. Thus, we use DIA, for diagrams not Many of you (Stephen Seiler, Stuart Gallo- based on numbers (e.g., flow charts) and way, Felipe Carpes, Catherine Bacon, Conrad , for making graphs from statistical Earnest, Paul Montgomery) sang the praises of and other data. Gnuplot can also very easily be Prism and pointed out that its graphs come used in batch mode, thus producing lots of apart in Powerpoint. I downloaded a trial ver- graphs reading a list of variables, a list of sub- sion to see, and yes, it's true. In Powerpoint jects, and so on. Both applications are free and 2003 the formatting is retained perfectly, can save graphics in formats that Powerpoint whereas in Powerpoint 2007 lines lose the can read." thickness you have given them in Prism, but Filipe Carpe recommended Microcal Origin. that's not a major problem. I also checked out He said Powerpoint will edit its graphs. The Pro Prism's statistical capabilities: it's more power- version is US$800. ful than Excel but not as powerful as Statistica. I haven't evaluated Stata, but from what I A problem with GraphPad is cost: around could find at its website, the graphs are editable $US450 for a single academic license. I won't in Powerpoint. Price depends on the version be buying one yet. you want: US$600 to $1600+. Prism will open an Excel spreadsheet of data, Other Solutions and Non-solutions but any graphs in the spreadsheet get corrupted. I had hopes for the free OpenOffice suite, Prism is therefore an adjunct for Excel, not a which features a spreadsheet very similar to replacement. Excel 2003. It opens my spreadsheets, but the Solution #3: SigmaPlot graph symbols do not have outlines and, incred- Several people (Martin Buchheit, Mike Ham- ibly, there is no plain circle symbol. The lin, Yngve Sommervoll) are happy with the OpenOffice equivalent of Powerpoint is quite graphing and other capabilities of SigmaPlot. I klunky and too different from Powerpoint to be got conflicting advice about pasting into Pow- worth persevering with. There were also too erpoint, even from the SigmaPlot site, so I many bugs when transferring a slide to Power- downloaded a trial version, created a graph, point 2003 or 2007. pasted it into Powerpoint as a Windows meta- Conrad Earnest suggested Kalidegraph. It file, then took it apart without any problems. looks promising, because it's not too expensive Price: US$550. (US$140), but I searched the site without suc- cess for anything about ability to edit in Power-

Sportscience 15, i-iii, 2011 In Brief Page iii point. Conrad also pointed me to a comprehen- add-in for Excel that allows assembly of several sive list of graph software at Wikipedia. Kevin graphs into a single panel of graphs for publica- Short pointed out that Kalidegraph works on tion. Unfortunately he found that it does not PC and Mac, "though it's still not easy to show allow editing of the graphs. a student quickly how to use it and copying Finally, SAS and SPSS produce awful graphs images to other formats like PPT results in loss that can't even be taken apart–a disgrace for of image quality in our hands." such high-end expensive statistics packages. It James Best used ("very similar to takes years to learn SAS, but it's still worth the Origin graphing software") to generate a graph, effort, in spite of the graph problems: you can which he turned into a file (using a converter read in data from the messiest text files, and it program) that Powerpoint could open. I was has the most wonderfully powerful state-of-the- able to ungroup it to produce editable lines, but art analytical procedures. I routinely use Excel text became un-editable images. to create graphs from the text output of SAS. Ken Quarrie: "You might want to check out SPSS is limited to data only in the usual col- SCIDAVIS as a tool for plotting/charting and umn format. On the analysis side it now has a putting into documents. It is free for download. mixed model that will handle complex repeated It can read in ascii files and has quite an array measurements, but you have to use dummy of functions." I checked out the site and discov- variables to make it work properly–instructive ered that it is an offshoot from qtiplot. I was not for your understanding of linear models, but not encouraged when I found that the site was last recommended for routine use. See Tips and updated more than a year ago. tricks for SPSS if you have no choice. Alan Batterham suggested XL toolbox, a free

Updates: Sample Size, Clinical Inferences, Validity and Reliability Will G Hopkins, Sport and Recreation, AUT University, Auckland, New Zealand. Email. Sportscience 15, iii, 2011 (sportsci.org/2011/inbrief.htm#updates). Reviewer: Alan M Batterham, University of Teesside, Middlesbrough, UK. Published June 2011. ©2011 Sample Size. Updates for the spreadsheet tive) via simply increasing the level for the are detailed at the beginning of the article on confidence interval from the default of 90% that sample size. In summary, there is now a panel is used for mechanistic inferences, which there- for sample size when the dependent variables is by also become more conservative. There is a count (e.g., of injuries or of events in a game), also provision for controlling the error rate and the panel for single event outcomes now when there are several independent inferences, allows inclusion of smallest beneficial and simply by inserting the number of inferences. harmful effects as risk difference, odds ratio I have included cells showing a less con- and hazard ratio (in addition to the risk ratio servative approach based on the odds ratio of that was there originally). The article also has a benefit to harm, whereby an effect can be used bullet point for sample size in a pilot reliability study for a controlled trial. if odds of benefit outweigh the odds of harm by Clinical Inferences are now included in the more than 66. (See the article on clinical infer- controlled-trial spreadsheets (pre-post parallel- ences for more.) Reducing the threshold proba- groups, post-only crossover, pre-post crossover) bilities for harm and benefit does not increase and in the spreadsheet for comparing group this odds ratio, and I have been unable to devise means. The changes represent implementation a systematic approach to increasing the odds of the principles explained in the article on ratio for more conservative inferences. The clinical, practical and mechanistic inferences. odds ratio should therefore be used sparingly or Briefly, a clinical inference is based on the maybe only once in a study. notion that a treatment or other effect should be Some cells in the spreadsheets have extensive used on people only if the risk of harm is suffi- comments to guide their use. In particular, there ciently low (<0.5%, "most unlikely") and the is an explanation about adding up chances of chance of benefit is sufficiently high (>25%, harm and benefit with multiple inferences as a "possible"). The spreadsheet allows these more efficient alternative to reducing the thresholds to be made smaller (more conserva- thresholds of 0.5% and 25%. However, I sus-

Sportscience 15, i-iii, 2011 Sportscience In Brief Page 14 pect this approach will require familiarity with Validity and Reliability. I dealt with these magnitude-based inferences that will take re- measurement concepts in an article/slideshow searchers months or years to acquire, if they devoted to assessing athletes published here in bother at all. It's much easier to stay with null- 2004, but I have been using a more generic hypothesis testing and drop the alpha level, updated version for teaching these concepts to even if it means you end up with "no effects"! graduate students. The slideshow has a refer- This update has been a long time coming, be- ence list of various relevant articles, amongst cause the changes were complex and time con- which is something you shouldn't miss: the suming. If you spot any minor errors or format- Socratic dialogue on comparison of measures ting issues, please get back to me. (i.e., validity studies) published here in 2010. ————

Sportscience 13, 13-15, 2009