Scatterplot - Issues Limiting Y Axis and Labeling Data Points

[Richard Armour]

Hello all,

I am attempting to plot a scatter plot (twoway in v18) which identifies the OR with associated 95% CI over 5 time points. I am running into two issues which I am struggling to resolve.

Data is as follows:

* Example generated by -dataex-. For more info, type help dataex
clear
input str9 StudyID double(OR lCI uCI) long TimePoint
"72 hours"  8.21 1.77 38.18 5
"30 days"   5.07 2.32  7.82 3
"2 months " 5.76 3.25  8.26 1
"3 months " 5.42 2.75 10.69 2
"6 months " 1.14  .68   1.9 4
end
label values TimePoint TimePoint
label def TimePoint 1 "2 months", modify
label def TimePoint 2 "3 months", modify
label def TimePoint 3 "30 days", modify
label def TimePoint 4 "6 months", modify
label def TimePoint 5 "72 hours", modify

Issue 1:
As can be seen in the data, the 72 hours 95%CI is very large compared with 6 months (for example), making the output quite difficult to read.

I have attempted to limit the Y axis using the following codes but the graph produced does not limit the axis. I have attempted to insert the force command at various points around the yscale and ylabel command without success either.

twoway (scatter OR TimePoint, ms(Oh) mc (gs5)) (rcap lCI uCI TimePoint, lc(gs5)), yscale (range(0, 15))

twoway (scatter OR TimePoint, ms(Oh) mc (gs5)) (rcap lCI uCI TimePoint, lc(gs5)), ylabel(0 1 5 10)

Is there way to modify the y axis to either
a) limit the y axis and have the 95%CI as an arrow at the uCI of 72 hours or
b) modify the y axis so the 95%CI of the lower values is more easily viewed?

Issue 2:
When attempting to add labels to the point estimates of the OR using the mlabel command, I receive the following error:

command mlabel is unrecognized
r(199);

I am not sure what to do about this.

Thank you for your assistance,

Richard Armour

[Andrew Musau]

I have attempted to limit the Y axis using the following codes but the graph produced does not limit the axis.

Stata does not allow you to truncate the axis. If arbitrary truncation were allowed, I suspect it would be misused, leading to misleading graphics. However, you can expand the axis using -yscale(range())-. This thread may be helpful on how to proceed.

When attempting to add labels to the point estimates of the OR using the mlabel command, I receive the following error:

command mlabel is unrecognized
r(199);

-mlabel()- is an option of scatter, not a command. See

help scatter##marker_label_options

[Nick Cox]

As a further detail the time points in #1 are out of order as a result of a misguided encode.

[Nick Cox]

As often, using logarithmic scale is a good reaction to the occurrence of extremes. Also, why treat time point just as ordinal?

* Example generated by -dataex-. For more info, type help dataex
clear
input str9 StudyID double(OR lCI uCI) long TimePoint Days
"72 hours"  8.21 1.77 38.18 1  3
"30 days"   5.07 2.32  7.82 2  30
"2 months " 5.76 3.25  8.26 3  66
"3 months " 5.42 2.75 10.69 4  91
"6 months " 1.14  .68   1.9 5  183
end
label values TimePoint TimePoint
label def TimePoint 3 "2 months", modify
label def TimePoint 4 "3 months", modify
label def TimePoint 2 "30 days", modify
label def TimePoint 5 "6 months", modify
label def TimePoint 1 "72 hours", modify

scatter OR TimePoint || rcap lCI uCI TimePoint, ytitle(Odds ratio) yla(1 2 5 10 20 50) xla(1/5, valuelabel) ysc(log)