Albatross Plot - By() more than one variable

[Nate Tillern]

I am creating an albatross plot (basically a scatter of p-values vs sample size) of compiled literature. I have two dimensions over which I would like to differentiate the points in the plot. One is whether the paper is rated as "weak" or "strong" and the other is what the broad topic of the paper is, 1 of 5 options. Using the Albatross command with the by() option you can tell STATA to differentiate the points visually on the plot by one variable. It also has a color option, but color just seems to use the variable specified in the by() option.

I would ideally like to specify my rating variable in the by() option, and then have different colors differentiate the topic of the paper. I initially just tried this:

albatross SampleSize p_value direction, type(correlation) by(Rating) color(Topic)

However, the color option does not allow for a varlist or varname to be used with it.

I am hoping to learn either a way to work around this issue or if there is a module I can install that gives more power when creating plots and editing their appearance. Thank you!

[Nick Cox]

You don't explain the provenance of albatross.

You will only get different colours if the response variable is split into several variables. If albatross doesn't allow that on the fly, and I imagine that it would be documented if it did, no extra option will achieve this. You'll need different commands.

search separate

gives a clickable link to

SJ-5-4 gr0023 . . . . Stata tip 27: Classifying data points on scatter plots
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . N. J. Cox
Q4/05 SJ 5(4):604--606 (no commands)
tips for using separate, gray-scale gradation, and text
characters as class symbols to classify data points on
a scatter plot