"SR Millis" <[EMAIL PROTECTED]> wrote in message [EMAIL PROTECTED]">news:[EMAIL PROTECTED]... > I'm graphically-challenged and need help on the simple graphing/plotting > of confidence intervals. > > Specifically: I have 4 studies that each involve binary outcome in a > 2-group parallel design. I have calculated the differences in > proportions and their associated 95% CIs. I now want to graphically > display the point estimates and CIs together--like > > -----X----- > > --X-- > > ------X------ > > I there a simple way to produce summary graphs like this with Stata, > SAS, S-plus, SPSS, Excel, etc?
If you have access to one of the above, most stat packages have a charting function that supports overlaying error bars on barcharts or x-y scatter/point plots. Either there are actual error bar functions or there are line segment plotting/drawing functions that you can program to draw the CIs. It has been a while since I used SAS, but my recollection is that the MOVE and DRAW annotation commands can be used to update an underlying graphic by using a dataset that has the error bar specifications. S-Plus, for example, has the error.bar() function and there are libraries/packages for S-Plus such as Frank Harrell's "Hmisc" (see errbar() ) which offer new or enhanced versions of many base functions. Hmisc is a standard package now with S-Plus. Similar to S-Plus is R (http://www.r-project.org/), which I use and it too has similar functionality (including Hmisc) available such as the segments() or arrows() functions that can be used to add either horizontal or vertical error bars to a plot. Lastly, Excel does have the ability to plot error bars. If you generate the underlying graphic, you can click on a data point or series, right click to bring up the menu and select "Format Data Series". This will bring up a tabbed dialogue which should have either a Y error bar or X error bar tab. You can then define either constants or cell ranges that have the upper and lower bounds of your intervals. HTH. Marc . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
