Time series line graphs

[James Alexander]

Hi there

I am trying to set up a time series line graph to illustrate the growth rate of two variables. I have generated two new variables to show the yearly growth rate, then added these to a two way line graph. Please see code below

gen L_Prod = L.Avg_Prod
gen GR_Prod = 100*(Avg_Prod-L_Prod)/L_Prod

gen L_Import_Total = L.Import_Total
gen GR_Import = 100*(Import_Total-L_Import_Total)/L_Import_Total

label var GR_Import "Import value growth rate (%)"

label var GR_Prod "Productivity growth rate (%)"

twoway tsline GR_Prod, yaxis(1) || tsline GR_Import, yaxis(2)

The x axis is year but I want to change the range. I have tried to use xlabel but it doesn't recognise any years I put in. Eg I tried xlabel (2016 2017 ...) and also xlabel (2016(1)2021).

Each time I had an error message saying 2016 is an invalid name. Perhaps this is because I'm working with panel data as apposed to time series data? Is there any way around this?

Screenshot of graph is attached for reference.

Thanks in advance

[Nick Cox]

I can't reproduce your problem in the absence of a data example. With panel data, a recipe like

webuse grunfeld, clear 

gen GR_invest = 100* D.invest/L.invest 

tsline GR_invest

just produces a tangle of lines, one for each panel. But adding an option like

tsline GR_invest, xla(1936(2)1954)

works to change the axis labels as shown.

Note the simpler way to get growth rates in my first code block.

[Nick Cox]

Try taking out any space after xlabel and before its argument. So xlabel(2016(1)2021) not xlabel (2016(1)2021) noting also that xlabel(2016/2021) is equivalent.

[James Alexander]

Thanks a lot for your help Nick, that is much appreciated.

Your suggestions worked and I have attached an updated screenshot of the graph.

One minor issue is that there appears to be a gap on the x axis before the first label (2016). This isn't the end of the world but I was wondering if there's any way to remove this?

[Nick Cox]

You can try xsc(r(2016, 2021)) or -- if that does not work -- temporarily dropping 2015 from the dataset.

preserve 
drop if year == 2015 
* your unstated command 
restore

I recommend almost any graph scheme over the default s2color.