Formatting a ITSA graph

[Scott Rick]

Hi. I would like to create the graph on the left. My code below, however, creates the graph on the right. I am unable to take off the bits above and below the graph ("Intervention starts..." and "GEE model...."). I am also unable to remove the fitted value line and add the lfit line to this.

Any help in doing so will be very much appreciated.

*code

xtset con_id week_num

xtitsa week_doctele i.age_cat i.doc_desig i.gender_cat i.department_cat i.LVPexp_cat ///
week_num, single trperiod(78; 114) posttrend figure (legend(order(1 "observed" 2 "fitted values") size(small)) title("Hyderabad")             ///
ytitle("Number of consultations") ylabel(#4, labsize(small) angle(horizontal)) ///
xline(145) xlab(35 "(1 Jan '22 - 29 Jun '23)" 95 "(30 June, 2023 - 5 Mar, 2024)" 127 "(6 Mar, 2024 - 31 Jul, 2024)", noticks labsize (vsmall)) ///
xtick(`tickcall',  tlength(*2)) xline(78) xline(114) scheme(s1color))                 ///        
 replace 
 


*data

input long con_id float(week_num week_doctele age_cat) byte doc_desig float(gender_cat department_cat LVPexp_cat)
1 118  .03448276 2 1 1 6 1
1 119  .01724138 2 1 1 6 1
1 120  .01724138 2 1 1 6 1
1 121  .03448276 2 1 1 6 1
1 126  .03448276 2 1 1 6 1
1 128  .01754386 2 1 1 6 1
1 131  .05263158 2 1 1 6 1
1 135 .071428575 2 1 1 6 1
2   5  .12068965 2 3 2 8 1
2   6  .05172414 2 3 2 8 1
2   7  .05172414 2 3 2 8 1
2   8  .12068965 2 3 2 8 1
2   9   .1754386 2 3 2 8 1
2  11  .19298245 2 3 2 8 1
2  13  .24561404 2 3 2 8 1
2  15   .2105263 2 3 2 8 1
2  17  .22807017 2 3 2 8 1
2  19  .16949153 2 3 2 8 1
2  20  .15254237 2 3 2 8 1
2  21  .10169491 2 3 2 8 1
2  23  .22413793 2 3 2 8 1
2  24  .14035088 2 3 2 8 1
2  25   .1754386 2 3 2 8 1
2  26  .19298245 2 3 2 8 1
2  27  .07017544 2 3 2 8 1
2  28  .24561404 2 3 2 8 1
2  29  .05263158 2 3 2 8 1
2  30  .15789473 2 3 2 8 1
2  31  .13793103 2 3 2 8 1
2  32  .25862068 2 3 2 8 1
2  33   .0862069 2 3 2 8 1
2  34   .1551724 2 3 2 8 1
2  35   .1551724 2 3 2 8 1
2  36   .2413793 2 3 2 8 1
2  37  .10344828 2 3 2 8 1
2  38   .1551724 2 3 2 8 1
2  39  .13793103 2 3 2 8 1
2  40  .10344828 2 3 2 8 1
2  41  .10344828 2 3 2 8 1
2  42  .12068965 2 3 2 8 1
2  43   .1724138 2 3 2 8 1
2  44  .10526316 2 3 2 8 1
2  45  .12280702 2 3 2 8 1
2  46  .19298245 2 3 2 8 1
2  47  .12280702 2 3 2 8 1
2  48  .14035088 2 3 2 8 1
2  49  .05263158 2 3 2 8 1
2  50  .12280702 2 3 2 8 1
2  51   .2105263 2 3 2 8 1
2  52  .12280702 2 3 2 8 1
2  53   .1754386 2 3 2 8 1
2  54  .12280702 2 3 2 8 1
2  55   .2631579 2 3 2 8 1
2  57  .08064516 2 3 2 8 1
2  59   .2580645 2 3 2 8 1
2  61   .1935484 2 3 2 8 1
2  62  .06451613 2 3 2 8 1
2  63  .06451613 2 3 2 8 1
2  64  .12903225 2 3 2 8 1
2  65  .11290322 2 3 2 8 1
2  66  .09836066 2 3 2 8 1
2  67   .1147541 2 3 2 8 1
2  68  .13114753 2 3 2 8 1
2  69  .08196721 2 3 2 8 1
2  72  .21311475 2 3 2 8 1
2  73  .06557377 2 3 2 8 1
2  75   .1967213 2 3 2 8 1
2  76   .1147541 2 3 2 8 1
2  77  .09836066 2 3 2 8 1
2  78  .18032786 2 3 2 8 1
2  79  .18333334 2 3 2 8 1
2  81         .2 2 3 2 8 1
2  82        .15 2 3 2 8 1
2  83        .15 2 3 2 8 1
2  84  .18333334 2 3 2 8 1
2  85  .18644068 2 3 2 8 1
2  87    .220339 2 3 2 8 1
2  89   .3050847 2 3 2 8 1
2  90  .16949153 2 3 2 8 1
2  91  .20338982 2 3 2 8 1
2  93  .13559322 2 3 2 8 1
2  96    .220339 2 3 2 8 1
2  97    .220339 2 3 2 8 1
2  98  .18644068 2 3 2 8 1
2  99  .15254237 2 3 2 8 1
2 102  .16949153 2 3 2 8 1
2 103  .15254237 2 3 2 8 1
2 104  .20338982 2 3 2 8 1
2 109  .13793103 2 3 2 8 1
2 111   .1724138 2 3 2 8 1
2 113  .29310346 2 3 2 8 1
2 114  .13793103 2 3 2 8 1
2 116  .18965517 2 3 2 8 1
2 117  .20689656 2 3 2 8 1
2 118   .2413793 2 3 2 8 1
2 119   .1724138 2 3 2 8 1
2 120   .0862069 2 3 2 8 1
2 121  .13793103 2 3 2 8 1
2 122  .18965517 2 3 2 8 1
2 123  .18965517 2 3 2 8 1

[George Ford]

you might get more help with a dataex that will produce results.

[Scott Rick]

George Ford Hi George. I apologize, I thought the data sample above would suffice. My dataset has about 4,500 observations and I'm unsure how to provide a data sample or the entire dataset to produce results. Could you please guide me on this?

[George Ford]

not sure. tricky given large sample and convergence issues. maybe just a few observations per time period. find something that runs and gets you close to what you've got (visually).

[Scott Rick]

George Ford Hi George. Here's a sample that works with the code below to generates the graph in black. I'm looking to create the graph in green. I'd greatly appreciate any help on this.

xtset con_id week_num

xtitsa week_doctele, single trperiod(29; 47) posttrend figure (legend(order(1 "observed" 2 "fitted values") size(small)) title("Hyderabad")             ///
ytitle("Number of consultations") ylabel(#4, labsize(small) angle(horizontal)) ///
 xlab(20 "(1 Jan '22 - 29 Jun '23)" 35 "(30 June, 2023 - 5 Feb, 2024)" 65 "(6 Feb, 2024 - 31 Jul, 2024)", noticks labsize (vsmall)) ///
xtick(`tickcall',  tlength(*2)) xline(29) xline(47) scheme(s1color))                 ///        
 replace 

input long con_id float(week_num week_doctele age_cat gender_cat) byte doc_desig float(department_cat LVPexp_cat)
 5  1   .6153846 3 2 4 2 3
12  1   .7692308 2 2 1 7 1
 4  2  .07692308 2 1 5 2 2
10  3   .3846154 3 1 1 3 2
 4  4   .9230769 2 1 5 2 2
12  4   .9230769 2 2 1 7 1
10  5  .15384616 3 1 1 3 2
 5  6   .7692308 3 2 4 2 3
 1  8          1 3 2 2 7 2
 1  9  .53846157 3 2 2 7 2
 5 10   .9230769 3 2 4 2 3
11 10  .23076923 2 2 2 4 1
10 11   .8461539 3 1 1 3 2
13 11  1.6153846 3 1 3 7 3
 6 12  .53846157 2 1 5 6 2
10 12   .6153846 3 1 1 3 2
15 12   .8461539 2 2 2 2 2
 4 13  .53846157 2 1 5 2 2
10 13  .53846157 3 1 1 3 2
12 13   2.692308 2 2 1 7 1
 4 14   .9230769 2 1 5 2 2
 8 16   .4615385 2 1 2 4 1
 4 17  1.5384616 2 1 5 2 2
12 17  1.5384616 2 2 1 7 1
 1 18   .7692308 3 2 2 7 2
12 18   .8461539 2 2 1 7 1
13 18  1.6923077 3 1 3 7 3
 1 19   .7692308 3 2 2 7 2
 3 19  1.6153846 2 1 1 2 1
 8 19   .6923077 2 1 2 4 1
 1 21   .9230769 3 2 2 7 2
 3 21  1.3076923 2 1 1 2 1
13 21  1.3846154 3 1 3 7 3
 5 22   .5833333 3 2 4 2 3
 1 23   .9166667 3 2 2 7 2
 9 24  .16666667 3 1 2 5 1
 6 25          4 2 1 5 6 2
 6 26        1.5 2 1 5 6 2
 1 27        1.5 3 2 2 7 2
 4 27   .8333333 2 1 5 2 2
11 28   .8333333 2 2 2 4 1
 1 29  1.5833334 3 2 2 7 2
 3 29  1.6666666 2 1 1 2 1
 8 31   .6428571 2 1 2 4 1
 8 32   .7857143 2 1 2 4 1
12 32  1.3571428 2 2 1 7 1
11 33         .5 2 2 2 4 1
 1 34          1 3 2 2 7 2
12 35   .7142857 2 2 1 7 1
15 35  1.2142857 2 2 2 2 2
15 36   .6428571 2 2 2 2 2
 1 38         .5 3 2 2 7 2
11 38  1.1428572 2 2 2 4 1
 8 39   .3571429 2 1 2 4 1
 8 40   .3571429 2 1 2 4 1
 9 42   .4285714 3 1 2 5 1
11 43  1.1428572 2 2 2 4 1
 8 44   .5714286 2 1 2 4 1
11 45  1.1428572 2 2 2 4 1
11 47  1.2857143 2 2 2 4 1
15 48  1.4285715 2 2 2 2 2
 6 49   .7857143 2 1 5 6 2
 4 50   .4285714 2 1 5 2 2
 4 51   .8571429 2 1 5 2 2
 3 53   .6428571 2 1 1 2 1
 9 56   .2857143 3 1 2 5 1
 6 58   .7857143 2 1 5 6 2
 4 59   .7857143 2 1 5 2 2
15 59   .9285714 2 2 2 2 2
 1 60  1.1428572 3 2 2 7 2
 4 60        1.5 2 1 5 2 2
 5 61   .4285714 3 2 4 2 3
15 63   .9285714 2 2 2 2 2
 8 64   .2142857 2 1 2 4 1
12 65  1.4285715 2 2 1 7 1
12 66  1.2142857 2 2 1 7 1
15 66   .7857143 2 2 2 2 2
11 67         .5 2 2 2 4 1
 4 68 .071428575 2 1 5 2 2
 4 69   .2857143 2 1 5 2 2
12 69  1.5714285 2 2 1 7 1
 3 70   .4285714 2 1 1 2 1
15 74          1 2 2 2 2 2
16 74 .071428575 2 1 1 2 1
 4 75   .6428571 2 1 5 2 2
 8 75   .7857143 2 1 2 4 1
 8 76   .4285714 2 1 2 4 1
15 76   .6428571 2 2 2 2 2
 1 77   .5714286 3 2 2 7 2
 4 77   .3571429 2 1 5 2 2
 4 78   .7142857 2 1 5 2 2
15 78   .8571429 2 2 2 2 2
 5 79         .5 3 2 4 2 3
15 79   .2857143 2 2 2 2 2
13 80   .4285714 3 1 3 7 3
13 82  1.0714285 3 1 3 7 3
 3 83   .2142857 2 1 1 2 1
13 83  1.3571428 3 1 3 7 3
 1 84   .8571429 3 2 2 7 2
 3 84   .7857143 2 1 1 2 1

[Andrew Musau]

note("") subtitle("")

should remove the note and subtitle. This is a simple scatter-plot, why not just use twoway to create the graph that you need? I am unable to do this for you for now. Note that xtitsa is from SSC, as you are asked to explain (FAQ Advice #12).

[Scott Rick]

Andrew Musau Thanks. xtitsa uses twoway options in the form of figure[(twoway_options)]. I just haven't been able to figure them out to get the graph I want.

[George Ford]

You should contact Ariel Linden at lindenconsulting. It's his code and maybe he can figure it out. His paper shows linear lines, but I can't sort it out.

Here's what Andrew was alluding to.

twoway lfit week_doctele week_num if week_num<=28 || lfit week_doctele week_num if inrange(week_num,29,46) || lfit week_doctele week_num if week_num>=47 || scatter week_doctele week_num , xline(29 47)