Problem with twoway scatter: Why do the values remain continuously at y=0 on the graph?

[Michael Duarte Goncalves]

I'd like to post a variable on the x-axis, ranging from 0 to 90000. However, the values are fixed and are not necessarily continuous (it starts at 0, then continues at 1, jumps to a non-sequential value, etc). I wanted to make a graph with y showing the number of observations per id depending on whether the power_p variable is 0, 1, ...

Next, I'd like to plot this graph by product, i.e. the categorical variable "product_classification2" (1,2,3,4). This variable is nominal, not cardinal. The "order" is not important.

The "total3" variable represents a count the id's, by "power_p". "Power_p3" equals 1 if "power_p" is 0, 2 if 3, and so on.

* Example generated by -dataex-. For more info, type help dataex
clear
input float id_numerical long(power_p power_p3) float total3 long product_classification2
 167391 0   2 21 1
 606274 0   2 21 2
 463980 0   2 21 1
  12089 0   2 21 4
 463980 0   2 21 1
 332979 0   2 21 1
 606274 0   2 21 2
 196748 0   2 21 1
 573193 0   2 21 2
 727027 0   2 21 1
  12089 0   2 21 4
 547007 0   2 21 2
 332979 0   2 21 1
  37068 0   2 21 2
  37068 0   2 21 2
 196748 0   2 21 1
 606274 0   2 21 2
 467938 0   2 21 2
  37068 0   2 21 2
 606274 0   2 21 2
  37068 0   2 21 2
 261637 1   3 50 2
 373646 1   3 50 1
1244020 1   3 50 2
 363069 1   3 50 2
  66665 1   3 50 2
 775905 1   3 50 2
 369693 1   3 50 2
 131820 1   3 50 2
 878192 1   3 50 2
 304150 1   3 50 1
1095955 1   3 50 1
 595716 1   3 50 2
 347416 1   3 50 1
 785459 1   3 50 2
 410882 1   3 50 2
1064514 1   3 50 2
 976872 1   3 50 2
 410882 1   3 50 2
  88765 1   3 50 1
 223320 1   3 50 2
 304150 1   3 50 1
1236813 1   3 50 2
 347416 1   3 50 1
 347416 1   3 50 1
1095955 1   3 50 1
 604293 1   3 50 2
 870793 1   3 50 2
 347416 1   3 50 1
1076215 1   3 50 2
1192145 1   3 50 1
1085856 1   3 50 2
 922230 1   3 50 1
1095955 1   3 50 1
 870891 1   3 50 2
 803598 1   3 50 1
 347416 1   3 50 1
1095955 1   3 50 1
1008453 1   3 50 2
 853336 1   3 50 1
1095955 1   3 50 1
 151802 1   3 50 2
 853336 1   3 50 1
   6117 1   3 50 2
 630905 1   3 50 2
 542584 1   3 50 2
 846853 1   3 50 2
 527149 1   3 50 2
 803598 1   3 50 1
 655872 1   3 50 2
 373646 1   3 50 1
 602225 2 463  2 1
 602225 2 463  2 1
 924145 3 722 13 1
 754578 3 722 13 1
 754578 3 722 13 1
  88766 3 722 13 1
  22159 3 722 13 1
1074655 3 722 13 1
 832257 3 722 13 1
 832257 3 722 13 1
 313726 3 722 13 1
 924145 3 722 13 1
  88766 3 722 13 1
1074655 3 722 13 1
 313726 3 722 13 1
 231538 4 899 14 2
   6946 4 899 14 1
 317723 4 899 14 2
 694927 4 899 14 1
1070575 4 899 14 1
 317723 4 899 14 2
1093714 4 899 14 2
1093714 4 899 14 2
1070575 4 899 14 1
 231538 4 899 14 2
   6946 4 899 14 1
 816495 4 899 14 1
 694927 4 899 14 1
 816495 4 899 14 1
end
label values power_p3 power_p3
label def power_p3 2 "0", modify
label def power_p3 3 "1", modify
label def power_p3 463 "2", modify
label def power_p3 722 "3", modify
label def power_p3 899 "4", modify
label values product_classification2 product_classification2
label def product_classification2 1 "Clasico", modify
label def product_classification2 2 "Indexado", modify
label def product_classification2 4 "Tarifa Justa", modify

Here is the graph that I obtained:

Best,

Michael

[Nick Cox]

Please show the command you used to get the graph in #1,

[Michael Duarte Goncalves]

This is the most basic one that I used:

graph twoway scatter total3 power_p3

I then used the ,by option but obtain weird things.

Basically, this is the kind of graphs that I want, by category product_classification2 (contained 1,2,3,4 for each category):

Thank you for your help Nick Cox

[Nick Cox]

Thanks for further detail. I can't follow all that you want here, but

* I wouldn't use a scatter plot to show frequencies

* I don't gather that your category scale makes smoothing helpful, or even valid

* You seem surprised, yet at the same time expectant, that a few categories are common, but many categories are not , and so that this will show up on a graph.

* The value label definitions seem backward to me.

All that said, this may help. Perhaps the major point is use of a square root scale to dampen the contrast of common and rare categories.

At the same time, I fear that the full version of this with all the data will still be a mess and that you might be better off with focus on the most common categories, whatever they are.

* Example generated by -dataex-. For more info, type help dataex
clear
input float id_numerical long(power_p power_p3) float total3 long product_classification2
 167391 0   2 21 1
 606274 0   2 21 2
 463980 0   2 21 1
  12089 0   2 21 4
 463980 0   2 21 1
 332979 0   2 21 1
 606274 0   2 21 2
 196748 0   2 21 1
 573193 0   2 21 2
 727027 0   2 21 1
  12089 0   2 21 4
 547007 0   2 21 2
 332979 0   2 21 1
  37068 0   2 21 2
  37068 0   2 21 2
 196748 0   2 21 1
 606274 0   2 21 2
 467938 0   2 21 2
  37068 0   2 21 2
 606274 0   2 21 2
  37068 0   2 21 2
 261637 1   3 50 2
 373646 1   3 50 1
1244020 1   3 50 2
 363069 1   3 50 2
  66665 1   3 50 2
 775905 1   3 50 2
 369693 1   3 50 2
 131820 1   3 50 2
 878192 1   3 50 2
 304150 1   3 50 1
1095955 1   3 50 1
 595716 1   3 50 2
 347416 1   3 50 1
 785459 1   3 50 2
 410882 1   3 50 2
1064514 1   3 50 2
 976872 1   3 50 2
 410882 1   3 50 2
  88765 1   3 50 1
 223320 1   3 50 2
 304150 1   3 50 1
1236813 1   3 50 2
 347416 1   3 50 1
 347416 1   3 50 1
1095955 1   3 50 1
 604293 1   3 50 2
 870793 1   3 50 2
 347416 1   3 50 1
1076215 1   3 50 2
1192145 1   3 50 1
1085856 1   3 50 2
 922230 1   3 50 1
1095955 1   3 50 1
 870891 1   3 50 2
 803598 1   3 50 1
 347416 1   3 50 1
1095955 1   3 50 1
1008453 1   3 50 2
 853336 1   3 50 1
1095955 1   3 50 1
 151802 1   3 50 2
 853336 1   3 50 1
   6117 1   3 50 2
 630905 1   3 50 2
 542584 1   3 50 2
 846853 1   3 50 2
 527149 1   3 50 2
 803598 1   3 50 1
 655872 1   3 50 2
 373646 1   3 50 1
 602225 2 463  2 1
 602225 2 463  2 1
 924145 3 722 13 1
 754578 3 722 13 1
 754578 3 722 13 1
  88766 3 722 13 1
  22159 3 722 13 1
1074655 3 722 13 1
 832257 3 722 13 1
 832257 3 722 13 1
 313726 3 722 13 1
 924145 3 722 13 1
  88766 3 722 13 1
1074655 3 722 13 1
 313726 3 722 13 1
 231538 4 899 14 2
   6946 4 899 14 1
 317723 4 899 14 2
 694927 4 899 14 1
1070575 4 899 14 1
 317723 4 899 14 2
1093714 4 899 14 2
1093714 4 899 14 2
1070575 4 899 14 1
 231538 4 899 14 2
   6946 4 899 14 1
 816495 4 899 14 1
 694927 4 899 14 1
 816495 4 899 14 1
 end 
 
 label values power_p3 power_p3
label def power_p3 2 "0", modify
label def power_p3 3 "1", modify
label def power_p3 463 "2", modify
label def power_p3 722 "3", modify
label def power_p3 899 "4", modify
label values product_classification2 product_classification2
label def product_classification2 1 "Clasico", modify
label def product_classification2 2 "Indexado", modify
label def product_classification2 4 "Tarifa Justa", modify

levelsof power_p, local(xvals)

foreach x of local xvals { 
    su power_p3 if power_p == `x', meanonly 
    local xla `xla' `x' "`r(mean)'"
}

* install from Stata Journal 
mylabels 0 25 100 225 400 625 900 1225, myscale(sqrt(@)) local(yla) 


spikeplot power_p [fw=total3], lw(thick) root xla(`xla') yla(`yla') ytitle(Frequency (root scale)) by(product, note(""))

[Michael Duarte Goncalves]

Good afternoon Nick Cox,

Thank you so much for the tips provided.
You are totally right.


All the best,

Michael