There is less information in 5 parameters than in all 100 data points? How surprising.
Of course there is less information but that’s the point of statistics to create few statistical values out of many. Still box plots are way better than only showing the median or even the mean.
Uh, I don’t know about the rest of you all, but when I make variability or categorical X vs continuous Y plots in JMP I simply setup my preferences for box plots AND showing points. (and mean diamonds, outliers, connect cell means to show trends on ordinal variables, etc.)
Is doing both like this not the normal way to display data? Then I can code by color or symbol for other relevant variables to increase the amount of information displayed.
Good box plots should also have accompanying histograms which would also illustrate the point from OP.
Moral of the story – box plots aren’t problematic, *poorly made* box plots are problematic.
2 comments
There is less information in 5 parameters than in all 100 data points? How surprising.
Of course there is less information but that’s the point of statistics to create few statistical values out of many. Still box plots are way better than only showing the median or even the mean.
Uh, I don’t know about the rest of you all, but when I make variability or categorical X vs continuous Y plots in JMP I simply setup my preferences for box plots AND showing points. (and mean diamonds, outliers, connect cell means to show trends on ordinal variables, etc.)
Is doing both like this not the normal way to display data? Then I can code by color or symbol for other relevant variables to increase the amount of information displayed.
Good box plots should also have accompanying histograms which would also illustrate the point from OP.
Moral of the story – box plots aren’t problematic, *poorly made* box plots are problematic.
Comments are closed.