Dear all,
In Stata, we can easily use xtline to plot the graph of panel data.
webuse grunfeld,clear
xtline invest
In Julia, How can I plot this graph?
julia> using RDatasets
julia> using DataFrames
julia> df = dataset("plm", "Grunfeld");
julia> using DataFrames
julia> using DataFramesMeta
julia> using RDatasets
julia> df = dataset("plm", "Grunfeld");
julia> gd = groupby(df, :Firm);
julia> p = plot(layout = 10)
julia> for i in 1:gd.ngroups
data = @where(df, :Firm .== i)
@df data plot!(p[i], :Year, :Inv)
end
julia> display(p)
This is maybe not the best way to plot this graph.
AlgebraOfGraphics is pretty good for this.
using RDatasets
using DataFrames
using AlgebraOfGraphics, CairoMakie
df = dataset("plm", "Grunfeld");
plt = data(df) * mapping(:Year, :Inv, layout=:Firm => nonnumeric) * visual(Lines)
draw(plt)
:Firm => nonnumeric is required because the Firm column is an integer; otherwise, it would just be :Firm.
1 Like
Thanks so much. It is a better way to visualize panel data.
It seems not work.
julia> using RDatasets
julia> using DataFrames
julia> using AlgebraOfGraphics, CairoMakie
[ Info: Precompiling AlgebraOfGraphics [cbdf2221-f076-402e-a563-3d30da359d67]
julia> df = dataset("plm", "Grunfeld");
julia> plt = data(df) * mapping(:Year, :Inv, layout=:Firm => nonnumeric) * visual(Lines)
ERROR: UndefVarError: nonnumeric not defined
Stacktrace:
[1] top-level scope
@ REPL[26]:1
Hmm, you may have an out-of-date version of AlgebraOfGraphics and/or Julia. It looks like nonnumeric was introduced in v0.4 of AlgebraOfGraphics.
You can check the package version like
julia> import Pkg; Pkg.status()
Status `C:\Users\evan\.julia\environments\v1.7\Project.toml`
[cbdf2221] AlgebraOfGraphics v0.6.7
[6e4b80f9] BenchmarkTools v1.3.1
[13f3f980] CairoMakie v0.8.3
[324d7699] CategoricalArrays v0.10.6
[861a8166] Combinatorics v1.0.2
[a93c6f00] DataFrames v1.3.4
[864edb3b] DataStructures v0.18.13
[abce61dc] Decimals v0.4.1
[f6369f11] ForwardDiff v0.10.30
[6fe1bfb0] OffsetArrays v1.12.1
[27ebfcd6] Primes v0.5.2
[ce6b1742] RDatasets v0.7.7
[a759f4b9] TimerOutputs v0.5.19
You could try updating with julia> import Pkg; Pkg.update() but this updates all packages, which may break other things.
nonnumeric used to be called categorical according to the release notes, so you could also try
plt = data(df) * mapping(:Year, :Inv, layout=:Firm => categorical) * visual(Lines)
If that still doesnโt work, you can manually convert the Firm column to a categorical array.
using CategoricalArrays
df[!,:Firm] = categorical(df[:,:Firm]);
plt = data(df) * mapping(:Year, :Inv, layout=:Firm) * visual(Lines);
draw(plt)
(zhangdeliang) pkg> st
Status `~/Project.toml`
[cbdf2221] AlgebraOfGraphics v0.6.7
[13f3f980] CairoMakie v0.8.3
[e9467ef8] GLMakie v0.6.3
[7073ff75] IJulia v1.23.3
[49802e3a] ProgressBars v1.4.1
[ce6b1742] RDatasets v0.7.7
[3cdcf5f2] RecipesBase v1.2.1
[d1185830] SymbolicUtils v0.19.7
[fce5fe82] Turing v0.21.3
Hmm, does this raise the same error?
julia> AlgebraOfGraphics.nonnumeric(3)
AlgebraOfGraphics.NonNumeric{Int64}(3)
julia> using RDatasets
julia> using DataFrames
julia> using AlgebraOfGraphics, CairoMakie
julia> df = dataset("plm", "Grunfeld");
julia> plt = data(df) * mapping(:Year, :Inv, layout=:Firm => nonnumeric) * visual(Lines)
Layer(AlgebraOfGraphics.Visual(Lines, {}), 200ร5 DataFrameColumns
Row โ Firm Year Inv Value Capital
โ Int64 Int64 Float64 Float64 Float64
โโโโโโผโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
1 โ 1 1935 317.6 3078.5 2.8
2 โ 1 1936 391.8 4661.7 52.6
3 โ 1 1937 410.6 5387.1 156.9
4 โ 1 1938 257.7 2792.2 209.2
5 โ 1 1939 330.8 4313.2 203.4
6 โ 1 1940 461.2 4643.9 207.2
7 โ 1 1941 512.0 4551.2 255.2
8 โ 1 1942 448.0 3244.1 303.7
โฎ โ โฎ โฎ โฎ โฎ โฎ
194 โ 10 1948 5.66 68.0 5.38
195 โ 10 1949 4.21 71.24 7.39
196 โ 10 1950 3.42 69.05 8.74
197 โ 10 1951 4.67 83.04 9.07
198 โ 10 1952 6.0 74.42 9.93
199 โ 10 1953 6.53 63.51 11.68
200 โ 10 1954 5.12 58.12 14.33
185 rows omitted, Any[:Year, :Inv], {:layout = :Firm => AlgebraOfGraphics.nonnumeric})
julia> draw(plt)
Scene (800px, 600px):
0 Plots
22 Child Scenes:
โ Scene (800px, 600px)
โ Scene (800px, 600px)
โ Scene (800px, 600px)
โ Scene (800px, 600px)
โ Scene (800px, 600px)
โ Scene (800px, 600px)
โ Scene (800px, 600px)
โ Scene (800px, 600px)
โ Scene (800px, 600px)
โ Scene (800px, 600px)
โ Scene (800px, 600px)
โ Scene (800px, 600px)
โ Scene (800px, 600px)
โ Scene (800px, 600px)
โ Scene (800px, 600px)
โ Scene (800px, 600px)
โ Scene (800px, 600px)
โ Scene (800px, 600px)
โ Scene (800px, 600px)
โ Scene (800px, 600px)
โ Scene (800px, 600px)
โ Scene (800px, 600px)
So thereโs no more error, but itโs not drawing the plot now?
Sorry I am not very adept at the plotting ecosystemโฆ maybe someone with more experience could chime in.
Yes, there is no error now. But no graph.
Thanks.
julia> draw(plt)
Scene (800px, 600px):
0 Plots
22 Child Scenes:
โ Scene (800px, 600px)
โ Scene (800px, 600px)
โ Scene (800px, 600px)
โ Scene (800px, 600px)
โ Scene (800px, 600px)
โ Scene (800px, 600px)
โ Scene (800px, 600px)
โ Scene (800px, 600px)
โ Scene (800px, 600px)
โ Scene (800px, 600px)
โ Scene (800px, 600px)
โ Scene (800px, 600px)
โ Scene (800px, 600px)
โ Scene (800px, 600px)
โ Scene (800px, 600px)
โ Scene (800px, 600px)
โ Scene (800px, 600px)
โ Scene (800px, 600px)
โ Scene (800px, 600px)
โ Scene (800px, 600px)
โ Scene (800px, 600px)
โ Scene (800px, 600px)
Sorry, I missed the draw call. Are you executing this directly from the REPL? It should show in Pluto or VSCode. Otherwise you might want to have a look at ElectronDisplay.jl for plotting directly from the REPL.
Yes, I run it from REPL. I will try ElectronDisplay.jl. Thanks.
You can also just do fig = draw(plt) and then save(โmyplot.svgโ, fig) to save the figure to disk.
Yeah, youโll have to have some sort of display open. You can also use GLMakie for a desktop plot or WGLMakie for in-browser plots.
If you need to use CairoMakie, you can always save the plot and inspect it using an image viewer, i.e., save("filename.pdf", fig).
I raised a similar question here thought it was somehow ignored. 
It refers to PlotlyJS. I managed to produce a good output, but not without pain.
I now simplified a bit the code with a for loop to produce each graphic in the panel.