Bug in graph combine?

[Christopher Bratt]

I am trying to use graph combine in Stata 18, without success as soon as I include one specific variable.

I have been able to identify the variable "age" as seemingly causing the error. A simple demonstration with two histograms:

. hist age, name(g1, replace)
(bin=37, start=34, width=.89189189)

. hist age, name(g2, replace)
(bin=37, start=34, width=.89189189)

. graph combine g1 g2
too few quotes
r(132);

I can't see anything wrong with the variable age.

Problem is, -graph combine- works nicely with the data produced by dataex below.

. des age

Variable      Storage   Display    Value
    name         type    format    label      Variable label
---------------------------------------------------------------------------------------------------------------------------------------------
age             byte    %11.0f                Hva er din alder?

.
end of do-file

. do /var/folders/33/21gh5ql91ts2sfl0y9d_l1g80000gn/T/StataRun1684930937536.do

. dataex age


* Example generated by -dataex-. For more info, type help dataex
clear
input byte age
 .
 .
34
40
39
36
40
44
42
36
37
41
43
40
37
38
43
37
43
36
39
42
42
42
44
39
38
43
40
43
39
35
41
42
36
36
39
40
43
39
37
42
36
41
41
39
36
36
35
37
41
37
36
37
44
41
39
36
42
44
39
40
39
42
37
35
38
42
41
39
42
42
36
35
35
40
36
39
35
40
40
35
37
44
36
41
35
43
42
40
38
43
35
41
44
44
36
44
43
41
end

So there is something odd with this particular variable.

. codebook age

---------------------------------------------------------------------------------------------------------------------------------------------
age                                                                                                                        Hva er din alder?

---------------------------------------------------------------------------------------------------------------------------------------------

                  Type: Numeric (byte)

                 Range: [34,67]                       Units: 1
         Unique values: 34                        Missing .: 2/6,054

                  Mean: 50.5345
             Std. dev.: 8.45551

           Percentiles:     10%       25%       50%       75%       90%
                             38        44        51        57        62



. tab age, missing

 Hva er din |
    alder?
 |      Freq.     Percent        Cum.
------------+-----------------------------------
         34 |          1        0.02        0.02
         35 |        152        2.51        2.53
         36 |        170        2.81        5.34
         37 |        159        2.63        7.96
         38 |        168        2.78       10.74
         39 |        160        2.64       13.38
         40 |        170        2.81       16.19
         41 |        164        2.71       18.90
         42 |        167        2.76       21.66
         43 |        158        2.61       24.26
         44 |        167        2.76       27.02
         45 |        173        2.86       29.88
         46 |        193        3.19       33.07
         47 |        192        3.17       36.24
         48 |        194        3.20       39.44
         49 |        217        3.58       43.03
         50 |        256        4.23       47.26
         51 |        235        3.88       51.14
         52 |        268        4.43       55.57
         53 |        293        4.84       60.41
         54 |        292        4.82       65.23
         55 |        245        4.05       69.28
         56 |        218        3.60       72.88
         57 |        228        3.77       76.64
         58 |        229        3.78       80.43
         59 |        218        3.60       84.03
         60 |        157        2.59       86.62
         61 |        167        2.76       89.38
         62 |        122        2.02       91.39
         63 |        157        2.59       93.99
         64 |        110        1.82       95.80
         65 |        112        1.85       97.65
         66 |         61        1.01       98.66
         67 |         79        1.30       99.97
          . |          2        0.03      100.00
------------+-----------------------------------
      Total |      6,054      100.00

PS. If StataCorp would like the data, feel free to contact me.

[Hua Peng (StataCorp)]

I can not reproduce with a simulated dataset. Would you please either post the dataset here or send it to tech support. Please also include the following information, the output from

about
query graphics

and run:

 hist age
graph save hist.gph

post here or send hist.gph to us as well. Thanks.

[Nick Cox]

This works for me in Stata 18. Note the extra option discrete, needed but orthogonal to this thread.

* Example generated by -dataex-. For more info, type help dataex
clear
input byte age int freq
34   1
35 152
36 170
37 159
38 168
39 160
40 170
41 164
42 167
43 158
44 167
45 173
46 193
47 192
48 194
49 217
50 256
51 235
52 268
53 293
54 292
55 245
56 218
57 228
58 229
59 218
60 157
61 167
62 122
63 157
64 110
65 112
66  61
67  79
 .   2
end


expand freq 

label var age "Hva er din alder?"

hist age, name(G1, replace) discrete 

graph combine G1 G1

[Christopher Bratt]

. keep dv age

. save df_statalist, replace
file df_statalist.dta saved

. lowess dv age, jitter(10) name(g_1, replace)
. lowess dv age, jitter(10) name(g_2, replace)

. graph combine g_1 g_2
too few quotes
r(132);

[Nick Cox]

This problem I can reproduce. Remove the variable label and it goes away. But it's not the question mark that is the problem, and I can't see any exotic characters in the label.

[Hua Peng (StataCorp)]

Thanks, I can reproduce the issue. Interestingly, The error happens in Stata 17 and 16 as well (I did not go back further).

[Hua Peng (StataCorp)]

Ok, I know what is going on, for some reason, there is a new line character (ASCII char 10) at the end of the variable label for age in your dataset, i.e., the label is actually:

 "Hva er din alder?\n"

which causes graph combine to break.To verify:

. local name : var label age

. di strlen("`name'")
18

. di strlen("Hva er din alder")
16
 
. di strpos(`"`name'"', char(10))
18

. di tobytes("`name'")
\d072\d118\d097\d032\d101\d114\d032\d100\d105\d110\d032\d097\d108\d100\d101\d114\d063\d010

Relabeling the age varaible without the new line character at the end fixes the issue. BTW, that explains the header in tab age, missing in the first post.

[Nick Cox]

Good catch!

[Christopher Bratt]

Thanks a lot to both of you.

I can confirm that

label variable age "Age"

solves the problem. -graph combine- works as expected.

There are often problems with data delivered by others, this was a new one (and the data are from a well-establised polling organisation).


Actually, there was a giveaway in my -tab age- (see post #1 in this thread): The line break is clearly visible. But I wouldn't have understood the significance of that anyway.