Problem with the graph of xtline when using "i." command

[Jessie Ye]

Hi all,

I encountered some problems in graphing xtline diagrams. Here is my code:

preserve
collapse (mean)coverage, by (high planning finance performance year)
xtset year
xtline coverage , t(year) i(finance) overlay

NB: coverage is the y variable. finance is a category variable that includes 4 categories (finance has level 1.2.3.4, which is shown as line finance =1, finance =2...). The diagram will be like:

I am so confused why the diagram is like this, and only the line of finance level 4 looks normal.

Thanks a lot!

[Nick Cox]

You evidently have in your data set the variables high planning and performance and if the mean of coverage varies with those variables as well as with finance and year -- which seems inevitable in practice -- then what you asked for implies joining up different values for the covariates explicit on the graph.. Hence there are up and down connections as well as left to right connections.

As you don't want this graph, the question is what do you want to plot? The number of lines possible is the number of distinct combinations of four variables, it seems.

[Andrew Musau]

xtline coverage, t(year) i(finance) overlay

is equivalent to

xtset finance year
xtline coverage, overlay

The graph is consistent with multiple values of coverage in a particular year (exhibited by the vertical lines connecting points within the same year). I suspect that you are getting a repeated time values error after -xtset finance year-. Bottom line is that xtline is meant for panel data, so you need to somehow create a panel dataset from your variables (unique combinations of finance and year).

[Jessie Ye]

You evidently have in your data set the variables high planning and performance and if the mean of coverage varies with those variables as well as with finance and year -- which seems inevitable in practice -- then what you asked for implies joining up different values for the covariates explicit on the graph.. Hence there are up and down connections as well as left to right connections.

As you don't want this graph, the question is what do you want to plot? The number of lines possible is the number of distinct combinations of four variables, it seems.

Hi Nick，

Thanks a lot for your reply. What I try to plot is how the coverage rate of different levels (level 1.2.3.4) of assurance indicators (finance / performance / planning) changes over the years.

[Jessie Ye]

is equivalent to

xtset finance year
xtline coverage, overlay

The graph is consistent with multiple values of coverage in a particular year (exhibited by the vertical lines connecting points within the same year). I suspect that you are getting a repeated time values error after -xtset finance year-. Bottom line is that xtline is meant for panel data, so you need to somehow create a panel dataset from your variables (unique combinations of finance and year).

Hi Andrew,

yes, you are correct, I have repeated time error after xtset. but it seems inevitable in my dataset, see data below (collapse):

Here, finance 1 are appeared four times, because 4 medical organizations are rated as finance level 1, and their coverage rates in vaccines are exactly varied. So now I don't know how to avoid it.

year finance performance planning coverage
2005 1 1 2 96.4767
2005 1 3 1 96.775
2005 1 3 2 92.0667
2005 1 2 2 88.785
2005 2 3 3 90.73
2005 2 2 2 86.0205
2005 2 2 3 92.9025
2005 2 3 2 83.68
2005 3 1 1 88.8367
2005 3 2 3 93.0593
2005 3 2 2 86.936
2005 3 3 2 93.2767
2005 3 3 3 90.0925
2005 4 3 4 94.38
2006 1 1 2 96
2006 1 3 1 97.5
2006 1 3 2 94.6667
2006 1 2 2 87.25
2006 2 3 2 87
2006 2 3 3 96
2006 2 2 2 90.1278
2006 2 2 3 90.25
2006 3 2 3 95.0533
2006 3 2 2 88
2006 3 3 3 92.025
2006 3 1 1 90.6667
2006 3 3 2 94.6667
2006 4 3 4 94
2007 1 2 2 87.5
2007 1 3 2 93
2007 1 3 1 97.5
2007 1 1 2 96
2007 2 2 3 92.75
2007 2 3 3 94
2007 2 2 2 88.458
2007 2 3 2 97
2007 3 3 2 95.3333
2007 3 2 3 92.5333
2007 3 1 1 93
2007 3 2 2 87.4
2007 3 3 3 92.05
2007 4 3 4 95
2008 1 2 2 87.225
2008 1 3 1 98
2008 1 3 2 93.3333
2008 1 1 2 98
2008 2 3 2 94
2008 2 3 3 93.3333
2008 2 2 2 87.9954
2008 2 2 3 90.75
2008 3 3 2 94.6667
2008 3 2 3 93.1333
2008 3 3 3 92.2562
2008 3 2 2 85.2
2008 3 1 1 91
2008 4 3 4 95
2009 1 1 2 97.3667
2009 1 3 1 96.1
2009 1 3 2 94.7
2009 1 2 2 88.55
2009 2 2 2 87.748
2009 2 3 3 94.1333
2009 2 3 2 95.9
2009 2 2 3 91.35
2009 3 3 3 92.525
2009 3 1 1 94.7
2009 3 2 3 93.2733
2009 3 2 2 85.62
2009 3 3 2 95.7667
2009 4 3 4 95.1
2010 1 3 2 94.1333
2010 1 2 2 90.75
2010 1 1 2 98
2010 1 3 1 97.1
2010 2 2 2 90.4312
2010 2 2 3 94.3
2010 2 3 3 96.0667
2010 2 3 2 93.6
2010 3 2 2 86.76
2010 3 3 2 96.5
2010 3 3 3 94.2813
2010 3 2 3 95.2467
2010 3 1 1 95
2010 4 3 4 97.1
2011 1 3 2 95.3333
2011 1 1 2 98.1667
2011 1 2 2 92.875
2011 1 3 1 98.6
2011 2 2 2 91.2602
2011 2 3 2 95.2
2011 2 2 3 94.775
2011 2 3 3 95.3667
2011 3 2 2 93.32
2011 3 2 3 95.8333
2011 3 3 3 94.7375
2011 3 3 2 97.1
2011 3 1 1 94.4333
2011 4 3 4 98.6
2012 1 3 1 98.3
2012 1 1 2 98.3
2012 1 2 2 94.05
2012 1 3 2 95
2012 2 3 3 95.0333
2012 2 2 2 93.2146
2012 2 2 3 96.2
2012 2 3 2 95.9
2012 3 2 3 96.4267
2012 3 3 3 95.3375
2012 3 2 2 92.66
2012 3 3 2 96.9
2012 3 1 1 95.3
2012 4 3 4 98.3
2013 1 3 1 98.85
2013 1 1 2 98.4667
2013 1 2 2 93.3
2013 1 3 2 97.2
2013 2 3 3 97.6
2013 2 2 2 94.1154
2013 2 2 3 96.15
2013 2 3 2 96.8
2013 3 3 3 96.1312
2013 3 2 3 96.4467
2013 3 3 2 98.0333
2013 3 2 2 92.24
2013 3 1 1 95.3333
2013 4 3 4 99

[Andrew Musau]

It depends on what you want to show. If this is average coverage rate by finance level, then

preserve
collapse coverage, by(finance year)
xtset finance year
xtline coverage, overlay
restore

[Jessie Ye]

It depends on what you want to show. If this is average coverage rate by finance level, then

preserve
collapse coverage, by(finance year)
xtset finance year
xtline coverage, overlay
restore

Hi Andrew,

It works perfectly!!

Thanks so much for your help!

[Nick Cox]

Let's be clear. The route taken here -- preserve collapse xtset xtline restore -- can be got to work, but it is not essential.

This code is optimised for the data example. The full dataset evidently covers more years. I put in some work to kill the key (lose the legend) and label directly.

clear 
input year finance performance planning coverage
2005 1 1 2 96.4767
2005 1 3 1 96.775
2005 1 3 2 92.0667
2005 1 2 2 88.785
2005 2 3 3 90.73
2005 2 2 2 86.0205
2005 2 2 3 92.9025
2005 2 3 2 83.68
2005 3 1 1 88.8367
2005 3 2 3 93.0593
2005 3 2 2 86.936
2005 3 3 2 93.2767
2005 3 3 3 90.0925
2005 4 3 4 94.38
2006 1 1 2 96
2006 1 3 1 97.5
2006 1 3 2 94.6667
2006 1 2 2 87.25
2006 2 3 2 87
2006 2 3 3 96
2006 2 2 2 90.1278
2006 2 2 3 90.25
2006 3 2 3 95.0533
2006 3 2 2 88
2006 3 3 3 92.025
2006 3 1 1 90.6667
2006 3 3 2 94.6667
2006 4 3 4 94
2007 1 2 2 87.5
2007 1 3 2 93
2007 1 3 1 97.5
2007 1 1 2 96
2007 2 2 3 92.75
2007 2 3 3 94
2007 2 2 2 88.458
2007 2 3 2 97
2007 3 3 2 95.3333
2007 3 2 3 92.5333
2007 3 1 1 93
2007 3 2 2 87.4
2007 3 3 3 92.05
2007 4 3 4 95
2008 1 2 2 87.225
2008 1 3 1 98
2008 1 3 2 93.3333
2008 1 1 2 98
2008 2 3 2 94
2008 2 3 3 93.3333
2008 2 2 2 87.9954
2008 2 2 3 90.75
2008 3 3 2 94.6667
2008 3 2 3 93.1333
2008 3 3 3 92.2562
2008 3 2 2 85.2
2008 3 1 1 91
2008 4 3 4 95
2009 1 1 2 97.3667
2009 1 3 1 96.1
2009 1 3 2 94.7
2009 1 2 2 88.55
2009 2 2 2 87.748
2009 2 3 3 94.1333
2009 2 3 2 95.9
2009 2 2 3 91.35
2009 3 3 3 92.525
2009 3 1 1 94.7
2009 3 2 3 93.2733
2009 3 2 2 85.62
2009 3 3 2 95.7667
2009 4 3 4 95.1
2010 1 3 2 94.1333
2010 1 2 2 90.75
2010 1 1 2 98
2010 1 3 1 97.1
2010 2 2 2 90.4312
2010 2 2 3 94.3
2010 2 3 3 96.0667
2010 2 3 2 93.6
2010 3 2 2 86.76
2010 3 3 2 96.5
2010 3 3 3 94.2813
2010 3 2 3 95.2467
2010 3 1 1 95
2010 4 3 4 97.1
2011 1 3 2 95.3333
2011 1 1 2 98.1667
2011 1 2 2 92.875
2011 1 3 1 98.6
2011 2 2 2 91.2602
2011 2 3 2 95.2
2011 2 2 3 94.775
2011 2 3 3 95.3667
2011 3 2 2 93.32
2011 3 2 3 95.8333
2011 3 3 3 94.7375
2011 3 3 2 97.1
2011 3 1 1 94.4333
2011 4 3 4 98.6
2012 1 3 1 98.3
2012 1 1 2 98.3
2012 1 2 2 94.05
2012 1 3 2 95
2012 2 3 3 95.0333
2012 2 2 2 93.2146
2012 2 2 3 96.2
2012 2 3 2 95.9
2012 3 2 3 96.4267
2012 3 3 3 95.3375
2012 3 2 2 92.66
2012 3 3 2 96.9
2012 3 1 1 95.3
2012 4 3 4 98.3
2013 1 3 1 98.85
2013 1 1 2 98.4667
2013 1 2 2 93.3
2013 1 3 2 97.2
2013 2 3 3 97.6
2013 2 2 2 94.1154
2013 2 2 3 96.15
2013 2 3 2 96.8
2013 3 3 3 96.1312
2013 3 2 3 96.4467
2013 3 3 2 98.0333
2013 3 2 2 92.24
2013 3 1 1 95.3333
2013 4 3 4 99
end 

egen mean = mean(coverage), by(finance year)
egen tag = tag(finance year)
separate mean, by(finance) 

* ssc install mycolours  needed 
mycolours  

forval g = 1/4 { 
    local extra `extra' || scatter mean`g' year if year == 2013, ms(none) mla(finance) mlabc("`OK`g''") mlabsize(medium)
}

set scheme s1color 
line mean? year if tag, sort xla(2005(2)2013) lc("`OK1'" "`OK2'" "`OK3'" "`OK4'") ///
yla(88(2)98, ang(h)) xtitle("") legend(off) `extra' xsc(r(. 2013.2)) ytitle(Mean coverage by finance type)

[Jessie Ye]

Let's be clear. The route taken here -- preserve collapse xtset xtline restore -- can be got to work, but it is not essential.

This code is optimised for the data example. The full dataset evidently covers more years. I put in some work to kill the key (lose the legend) and label directly.

clear
input year finance performance planning coverage
2005 1 1 2 96.4767
2005 1 3 1 96.775
2005 1 3 2 92.0667
2005 1 2 2 88.785
2005 2 3 3 90.73
2005 2 2 2 86.0205
2005 2 2 3 92.9025
2005 2 3 2 83.68
2005 3 1 1 88.8367
2005 3 2 3 93.0593
2005 3 2 2 86.936
2005 3 3 2 93.2767
2005 3 3 3 90.0925
2005 4 3 4 94.38
2006 1 1 2 96
2006 1 3 1 97.5
2006 1 3 2 94.6667
2006 1 2 2 87.25
2006 2 3 2 87
2006 2 3 3 96
2006 2 2 2 90.1278
2006 2 2 3 90.25
2006 3 2 3 95.0533
2006 3 2 2 88
2006 3 3 3 92.025
2006 3 1 1 90.6667
2006 3 3 2 94.6667
2006 4 3 4 94
2007 1 2 2 87.5
2007 1 3 2 93
2007 1 3 1 97.5
2007 1 1 2 96
2007 2 2 3 92.75
2007 2 3 3 94
2007 2 2 2 88.458
2007 2 3 2 97
2007 3 3 2 95.3333
2007 3 2 3 92.5333
2007 3 1 1 93
2007 3 2 2 87.4
2007 3 3 3 92.05
2007 4 3 4 95
2008 1 2 2 87.225
2008 1 3 1 98
2008 1 3 2 93.3333
2008 1 1 2 98
2008 2 3 2 94
2008 2 3 3 93.3333
2008 2 2 2 87.9954
2008 2 2 3 90.75
2008 3 3 2 94.6667
2008 3 2 3 93.1333
2008 3 3 3 92.2562
2008 3 2 2 85.2
2008 3 1 1 91
2008 4 3 4 95
2009 1 1 2 97.3667
2009 1 3 1 96.1
2009 1 3 2 94.7
2009 1 2 2 88.55
2009 2 2 2 87.748
2009 2 3 3 94.1333
2009 2 3 2 95.9
2009 2 2 3 91.35
2009 3 3 3 92.525
2009 3 1 1 94.7
2009 3 2 3 93.2733
2009 3 2 2 85.62
2009 3 3 2 95.7667
2009 4 3 4 95.1
2010 1 3 2 94.1333
2010 1 2 2 90.75
2010 1 1 2 98
2010 1 3 1 97.1
2010 2 2 2 90.4312
2010 2 2 3 94.3
2010 2 3 3 96.0667
2010 2 3 2 93.6
2010 3 2 2 86.76
2010 3 3 2 96.5
2010 3 3 3 94.2813
2010 3 2 3 95.2467
2010 3 1 1 95
2010 4 3 4 97.1
2011 1 3 2 95.3333
2011 1 1 2 98.1667
2011 1 2 2 92.875
2011 1 3 1 98.6
2011 2 2 2 91.2602
2011 2 3 2 95.2
2011 2 2 3 94.775
2011 2 3 3 95.3667
2011 3 2 2 93.32
2011 3 2 3 95.8333
2011 3 3 3 94.7375
2011 3 3 2 97.1
2011 3 1 1 94.4333
2011 4 3 4 98.6
2012 1 3 1 98.3
2012 1 1 2 98.3
2012 1 2 2 94.05
2012 1 3 2 95
2012 2 3 3 95.0333
2012 2 2 2 93.2146
2012 2 2 3 96.2
2012 2 3 2 95.9
2012 3 2 3 96.4267
2012 3 3 3 95.3375
2012 3 2 2 92.66
2012 3 3 2 96.9
2012 3 1 1 95.3
2012 4 3 4 98.3
2013 1 3 1 98.85
2013 1 1 2 98.4667
2013 1 2 2 93.3
2013 1 3 2 97.2
2013 2 3 3 97.6
2013 2 2 2 94.1154
2013 2 2 3 96.15
2013 2 3 2 96.8
2013 3 3 3 96.1312
2013 3 2 3 96.4467
2013 3 3 2 98.0333
2013 3 2 2 92.24
2013 3 1 1 95.3333
2013 4 3 4 99
end

egen mean = mean(coverage), by(finance year)
egen tag = tag(finance year)
separate mean, by(finance)

* ssc install mycolours needed
mycolours

forval g = 1/4 {
local extra `extra' || scatter mean`g' year if year == 2013, ms(none) mla(finance) mlabc("`OK`g''") mlabsize(medium)
}

set scheme s1color
line mean? year if tag, sort xla(2005(2)2013) lc("`OK1'" "`OK2'" "`OK3'" "`OK4'") ///
yla(88(2)98, ang(h)) xtitle("") legend(off) `extra' xsc(r(. 2013.2)) ytitle(Mean coverage by finance type)

[ATTACH=CONFIG]n1601441[/ATTACH]

Hi Nick,

The diagram is perfectly working! Thanks so much!

By the way, I still have another question about the finance type. From the diagram, we see that the trend of type 1 and type 4 are similar. and when I run the command:

reg coverage i.year i.planning i.finance i.performance,r

the results show:

coverage	Coef.	St.Err.	t-value	p-value	[95% Conf	Interval]	Sig
Planning
D planning		Baseline
C planning	-1.088	.545	-2.00	.046	-2.158	-.018	**
B planning	1.815	.568	3.19	.001	.7	2.93	***
A planning	1.277	.617	2.07	.039	.067	2.488	**
Finance
D finance		Baseline
C finance	-2.253	.48	-4.69	.000	-3.195	-1.31	***
B finance	-2.326	.515	-4.52	.000	-3.335	-1.316	***
A finance		Omitted
Performance
D performance		Baseline
C performance	-3.117	.517	-6.03	.000	-4.131	-2.103	***
B performance	-1.68	.48	-3.50	.000	-2.623	-.738	***
*** p<.01, ** p<.05, * p<.1

So why the type 1 finance (A finance) is omitted by collinearity? is it because that the trend of type 1 (D finance) and type 4 (A finance) are similar (they are highly correlated)?

Thanks again!

[Nick Cox]

I'd ask #9 again in a new thread. The graphics question here is different. The test is whether a post matches the thread title.

But there is almost certainly an answer if you look at the data.

 
 graph matrix planning finance performance year