Graph box plot with loops using var label in ytitle

[paulfgross]

I am using Stata/IC 13.1 for Mac (64-bit Intel)

I want to produce 9 box plots of percentages of nine survey responses coded in nine variables pc_q21_`i'. (i=1…9)

The variable labels were defined earlier.For example,the coding for the label of the ninth variable q21_9 is as follows:

label variable q21_9 "Admission cause=none of above"
label define q21_9 9 "yes",modify
label values q21_9 q21_9

Two variables describe the state of residence ("state") and the wave of the survey("wave").There are eight states and three waves.

The following code sets up the loops for nine box plots of the proportions answering yes to each question, thanks to the graph "var label" hints in the new Cox-Newton 119 Stata Tips book, third edition (see page 263), and the egen code hint also due (I think) to Nick Cox in a recent Statalist response

forvalues i=1(1)9 {
egen pc_q21_`i' = mean(100 * (q21_`i' ==`i' & q21_`i' != . )), by(state wave)
tab pc_q21_`i' wave,missing row col
graph box pc_q21_`i' if tag, over(state) ytitle("`:var label q21_`i''") title("Percent of patients in PSS reporting", size(vsmall)) saving(AdmCause`i',replace)
}

Only in the ninth graph does the variable label (in this case "Admission cause=none of above") appear correctly in the ytitle. In the preceding eight graphs,only the relevant pc_q21_`i' appears.

Any ideas?

Paul Gross

[Nick Cox]

I am copying your code and formatting it as such to make it a little more readable.

 
label variable q21_9 "Admission cause=none of above"
label define q21_9 9 "yes",modify 
label values q21_9 q21_9

forvalues i=1(1)9 {
     egen pc_q21_`i' = mean(100 * (q21_`i' ==`i' & q21_`i' != . )), by(state wave)
     tab pc_q21_`i' wave,missing row col
     graph box pc_q21_`i' if tag, over(state) ytitle("`:var label q21_`i''") title("Percent of patients in PSS reporting", size(vsmall)) saving(AdmCause`i',replace)
}

It is hard to guess what the problem is here. Please show us the results of

 
describe q21_?

A secondary detail is that the second condition in

q21_`i' == `i'  & q21_`i' != .

is redundant as observations equal in turn to 1(1)9 cannot be missing. But that does no harm.

[paulfgross]

The suggestion is good.
Here is the coding that produced two of the variables:
. label variable q21_5 "Admission cause=DVT"

. label define q21_5 5 "yes",modify

. label values q21_5 q21_5

. tab q21_5,missing

Admission |
cause=DVT | Freq. Percent Cum.
------------+-----------------------------------
yes | 215 1.74 1.74
. | 12,109 98.26 100.00
------------+-----------------------------------
Total | 12,324 100.00

. label variable q21_6 "Admission cause=infection prev stay"

. label define q21_6 6 "yes",modify

. label values q21_6 q21_6

. tab q21_6,missing

Admission |
cause=infec |
tion prev |
stay | Freq. Percent Cum.
------------+-----------------------------------
yes | 215 1.74 1.74
. | 12,109 98.26 100.00
------------+-----------------------------------
Total | 12,324 100.00
So we seem to have the acceptable label.

I next ran Nick's suggestion, here is the result:

. describe q21_?

storage display value
variable name type format label variable label
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------
q21_1 byte %10.0g q21_1 Admission cause=fall
q21_2 byte %10.0g q21_2 Admission cause=problem with meds
q21_3 byte %10.0g q21_3 Admission cause=prob with hip replacement
q21_4 byte %10.0g q21_4 Admission cause=mental health issue
q21_5 byte %10.0g q21_5 Admission cause=DVT
q21_6 byte %10.0g q21_6 Admission cause=infection prev stay
q21_7 byte %10.0g q21_7 Admission cause=pressure sore prev stay
q21_8 byte %10.0g q21_8 Admission cause=other complication prev stay
q21_9 byte %10.0g q21_9 Admission cause=none of above

It seems that the variables are correctly lab led.

But when I run the series of codes that are intended to produce the nine box plots, I get the same result that produced my query to Statalist

Tiz a puzzlement as Yul Bruyner once opined.

Paul Gross

[Nick Cox]

That was Yul Brynner. We bald guys stick together. You're referring to "The King and I".

Poem from the late 1950s

Tough luck on poor Sid
His hair was a winner
Took three years to grow
Then along came Yul Brynner.

In terms of your Stata problem, I get no problems with this code

 
clear 
set obs 42 
set seed 2803 
forval j = 1/9 { 
    gen q21_`j' = 100 * runiform() 
}
label var q21_1 "Admission cause=fall"
label var q21_2 "Admission cause=problem with meds"
label var q21_3 "Admission cause=prob with hip replacement"
label var q21_4 "Admission cause=mental health issue"
label var q21_5 "Admission cause=DVT"
label var q21_6 "Admission cause=infection prev stay"
label var q21_7 "Admission cause=pressure sore prev stay"
label var q21_8 "Admission cause=other complication prev stay"
label var q21_9 "Admission cause=none of above"

forvalues i=1(1)9 {
      graph box q21_`i', ytitle("`:var label q21_`i''") title("Percent of patients in PSS reporting", size(vsmall)) saving(AdmCause`i',replace)
}

and I can't see that the differences in your code should make any difference. Are you sure that you are talking about the same version of the same data in the same Stata session? Sorry to descend to that level, but I am out of ideas otherwise.

[Konrad Zdeb]

Can I join discussion and ask for the additional two lines of code that would automatically save the graph in GPH and TIF (or similar format). I tried using the code below but did not achieve satisfactory results.

    graph save map_`mapvar' "C:\Stata Folder\Graphs\map_`mapvar'.gph", ///
        replace
        
    graph export map_`mapvar' "C:\Stata Folder\Exports\map_`mapvar'.tif", ///
        replace

Mapvar is defined as a simple local macro and passed to the spmap program. Spmap works flawlessly so I am guessing that I made mistake in the two lines of code concerned with export syntax.

[Nick Cox]

Someone asked me to give a proper reference. Here it is http://www.cartoons.ac.uk/record/GAA131588

[paulfgross]

Thanks to all.

Nick:It was the same data and version but I restarted the session.
The coding below gave me the different state averages, with all nine graphs titled correctly and stored in the Documents folder.

egen tag = tag(state wave)
forvalues i=1(1)9 {
egen pc_q21_`i' = mean(100 * (q21_`i' ==`i' & q21_`i' != .)), by(state wave)
tab pc_q21_`i' wave,missing row col
graph box pc_q21_`i' if tag, over(state) ytitle("`:var label q21_`i''") title("Percent of patients in PSS reporting", size(vsmall)) saving(AdmCause`i',replace)
}

However in the Graph window only the first and ninth graphs are shown in the vertical panel of graphs, the first titled "AdmCause1" and the last "Graph". To access the others I must open the Graph folder and click on them. A small inconvenience, but no longer a puzzlement to we who are follicly challenged but remember Yul with his surname now spelt correctly.

Paul Gross

[Nick Cox]

A simple tip is that if you name a graph, it remains open in its own window. So an extra option like

name(g1, replace)

ensures that. First time round, the replace is not needed, but if you like me you usually make a series of small changes it is a good idea for later versions.