Sequence Analysis - sqindexplot: change y-axis from n to proportion?

[JN Nielsen]

Dear Statalisters

I am using the *Sequence analysis with Stata' package to visualize body weight changes over time using

sqindexplot

For now, I can only shown 'n' on the y-axis, but I would like to have proportions instead.
Anyone who know how to do this? I have attached the plot and my code.
I have tried

sqpercentageplot

but it does not show how people changes weight status.

Kind regards,
Jannie

sqindexplot if Asample==1, color(green yellow orange red black) graphregion( color(white) ) legend (rows(5) size(vsmall)) ///
xtitle("Years of follow-up") xlabel(0 "0" 1 "2" 2 "4" 3 "6" 4 "8" 5 "10" 6 "12" 7 "14" 8 "16" 9 "18" 10 "20" 11 "22") ///
ytitle("n", height(10))

[Nick Cox]

Ulrich Kohler may be able to help.

search mylabels, sj

which may be part of a solution.

Otherwise, note the typo "Overweigth" -- which should be "Overweight".

[Ulrich Kohler]

There is nothing build-in in sqindexplot, but a variant of the following might be good enough as a workaround:

. sqset
. local n = r(imax)
. sqindexplot , ylab(5000 "`=5000/`n'" 1000 "`=10000/`n'")

[Nick Cox]

#3 is identical in spriit to what mylabels offers. Let's suppose that the number of people behind #1 is 13500.

mylabels creates a local macro you can use in your graph command.

. mylabels 0 0.25 0.5 0.75 1, myscale(@*13500) local(yla)
0 "0" 3375 ".25" 6750 ".5" 10125 ".75" 13500 "1"

In practice l like to see something like 0 0.25 0.5 0.75 1 and would edit in extra leading zeros as explicit characters.

[JN Nielsen]

Many thanks for your replies Ulrich Kohler and Nick Cox
Honestly, Im (still) not great with macros. I could not make Ulrich's code work.
With Nick's I read your article 'Speaking Stata...' and tried following:

scc install mylabels
mylabels 0.0 0.25 0.5 0.75 1, myscale(@*13401) local(yla)
0 "0" 3350 ".25" 6700 ".5" 10051 ".75" 13401 "1"
sqindexplot if Asample==1, color(green yellow orange red black) graphregion( color(white) ) legend (rows(5) size(vsmall)) ///
xtitle("Years of follow-up") xlabel(0 "0" 1 "2" 2 "4" 3 "6" 4 "8" 5 "10" 6 "12" 7 "14" 8 "16" 9 "18" 10 "20" 11 "22") ///
ytitle("Proportion", height(10)) ylabel(‘yla’)

and

sqindexplot if Asample==1, color(green yellow orange red black) graphregion( color(white) ) legend (rows(5) size(vsmall)) ///
xtitle("Years of follow-up") xlabel(0 "0" 1 "2" 2 "4" 3 "6" 4 "8" 5 "10" 6 "12" 7 "14" 8 "16" 9 "18" 10 "20" 11 "22") ///
ytitle("Proportion") ylabel(‘yla’) yscale((@))

However, for both I get the error 'invalid label specifier, : ‘yla’:'

Two last questions:
- can sq work with survey weighted data? (svy command)
- can sq calculate 'entropy'?

Kind regards,
Jannie

[Nick Cox]

First off, as implied by #2 mylabels should be referenced as coming from the Stata Journal and should in principle be installed from there. In practice, using ssc to install should be fine in this instance.

But there is no command scc to do that -- that should be ssc.

There may be other typos in what you post but it looks as if you're using the wrong quotation marks,

ylabel(`yla')

Confusingly, no doubt, the correct right quotation marks display as a vertical mark in many fonts.

I would be surprised if yscale((@)) worked. to do anything.

The other questions are for Ulrich Kohler

[Ulrich Kohler]

Nick Cox commented on your questions regarding the error message you got from Stata. I have nothing to add to Nick's answer. But here are the answers to the two other questions:

(1) Can sq work with survey weighted data? (svy command)

No. The SQ-Ados are merely a technique to classify the sequences observed in the data at hand. Consider the SQ-Ados just as a kind of -generate-/-replace- command for the construction of a very specific data management task. The concept of sampling weights, standard error or significance tests are thus meaningless.

Having said that you can, of course, apply -svy- concepts when you analyse the results of the SQ-Ados. For example, if you created clusters of sequences with -sqom- and subsequent cluster analysis, you calculate the standard error of the proportions of the clusters the usual way, and you can also do significance tests on differences in proportions and means. Perhaps this is what you have in mind.


(2) can sq calculate 'entropy'?

No, (but yes, perhaps). I haven't implemented any of the entropy-measures proposed for sequences. However, -sqom- has the option -sadi()-, which is a call to Brendan Halpins C-plugin for sequence analysis that takes advantage of the settings of the SQ-Ados. If this plugin offers some entropy measures it may run with -sqom, sadi(XXX)-. I haven't checked the functionality of sadi for years, and admittedly I don't even know if -sqom-'s interface to -sadi- still works. But it may be worth a try.

[JN Nielsen]

Once again - many thanks for your inputs! Makes sense with the survey weights. I will check out the sqom sadi option.

Nick Cox - my untrained (or just old) eyes did not catch the difference between the type of quotation marks - thanks for correcting.

But code is still not changing the y-axis. Issue may be the mylabels command. If I mark both lines before running the code, I getfollowing:

mylabels 0(0.25)1, myscale(@*13401) local(yla)
0 "0" 3350.25 ".25" 6700.5 ".5" 10050.75 ".75" 13401 "1"

0 "0" 3350 ".25" 6700 ".5" 10051 ".75" 13401 "1"

0 is not a valid command name
r(199);

However, if I only mark the first line (starting with 'mylabels'), code runs, but but y-scale does not change (see figure)

adoupdate mylabels

mylabels 0(0.25)1, myscale(@*13401) local(yla)
0 "0" 3350 ".25" 6700 ".5" 10051 ".75" 13401 "1"

sqindexplot if Asample==1, color(eltgreen green ebblue edkblue black) graphregion( color(white) ) legend (rows(5) size(vsmall)) ///
xtitle("Years of follow-up") xlabel(0 "0" 1 "2" 2 "4" 3 "6" 4 "8" 5 "10" 6 "12" 7 "14" 8 "16" 9 "18" 10 "20" 11 "22") ///
ytitle("Proportion") ylabel(`yla')

Kind regards,
Jannie

[Nick Cox]

Copying the mylabels output to the command line won't work -- it's not a command.

Referring to the local macro yla within a sqindexplot command should work.

Make sure that the command can see the local macro.

https://journals.sagepub.com/doi/pdf...36867X20931028

[JN Nielsen]

Thanks! its working!