# How can I show the mean in a groupedboxplot?

**URL:** https://discourse.julialang.org/t/how-can-i-show-the-mean-in-a-groupedboxplot/102980
**Category:** Visualization
**Tags:** plotting, dataframes, plots, statsplots
**Created:** [August 19, 2023, 11:05am UTC](https://discourse.julialang.org/t/how-can-i-show-the-mean-in-a-groupedboxplot/102980 "2023-08-19T11:05:11Z")
**Posts on this page:** 5
**Page:** 1

<div class="post-metadata">

### Author: ![bertulli](https://avatars.discourse-cdn.com/v4/letter/b/ecc23a/32.png) [@bertulli](https://discourse.julialang.org/u/bertulli)
#### Post date: [August 19, 2023, 11:05am UTC](https://discourse.julialang.org/t/how-can-i-show-the-mean-in-a-groupedboxplot/102980/1 "2023-08-19T11:05:11Z")

</div>

Hi all! I have this kind of DataFrame containing data:

```julia
julia> df_filtered = filter(:mnemonic => m -> m ∈ selected_instructions, df_temp)
1990×4 DataFrame
  Row │ mnemonic instruction forwarding value       
      │ String String Bool Quantity…   
──────┼───────────────────────────────────────────────────
    1 │ add add_r0_r2_#0 false 0.0829053 W
    2 │ add add_r0_r2_#1 false 0.0830969 W
    3 │ add add_r0_r2_#10 false 0.0831797 W
    4 │ add add_r0_r2_#100 false 0.083226 W
    5 │ add add_r0_r2_#101 false 0.0834463 W
    6 │ add add_r0_r2_#102 false 0.0834131 W
    7 │ add add_r0_r2_#103 false 0.0836367 W
    8 │ add add_r0_r2_#104 false 0.0830944 W
    9 │ add add_r0_r2_#105 false 0.0832901 W
   10 │ add add_r0_r2_#106 false 0.0832738 W
   11 │ add add_r0_r2_#107 false 0.0834709 W
   12 │ add add_r0_r2_#108 false 0.0833354 W
   13 │ add add_r0_r2_#109 false 0.0835491 W
   14 │ add add_r0_r2_#11 false 0.083368 W
  ⋮ │ ⋮ ⋮ ⋮ ⋮
 1977 │ mvn mvn_r6_r2 false 0.0863886 W
 1978 │ mvn mvn_r6_r3 false 0.0866079 W
 1979 │ mvn mvn_r6_r4 false 0.0864734 W
 1980 │ mvn mvn_r6_r5 false 0.0866562 W
 1981 │ mvn mvn_r6_r6 true 0.108087 W
 1982 │ mvn mvn_r6_r7 false 0.0870755 W
 1983 │ mvn mvn_r7_r0 false 0.0864812 W
 1984 │ mvn mvn_r7_r1 false 0.0866842 W
 1985 │ mvn mvn_r7_r2 false 0.0866603 W
 1986 │ mvn mvn_r7_r3 false 0.086884 W
 1987 │ mvn mvn_r7_r4 false 0.0867469 W
 1988 │ mvn mvn_r7_r5 false 0.0869432 W
 1989 │ mvn mvn_r7_r6 false 0.0869205 W
 1990 │ mvn mvn_r7_r7 true 0.108796 W
                                         1962 rows omitted

```

I want to produce a boxplot comparing the effect of the activation of the column “`forwarding`”. For now the best I could get is using `groupedboxplot()`, like this:

```julia
p = @df df_filtered groupedboxplot(:mnemonic, :value;
    group = :forwarding,
    # xlabel = "Instructions",
    ylabel = "Measured power",
    label = ["not active" "active"],
    bar_width = 0.7,
    left_margin = 5mm,
    bottom_margin = 3mm,
    size = (1000, 500),
    tickfontsize = 14,
    guidefontsize = 16
)

```

producing

 ![box_plot_selected](https://global.discourse-cdn.com/julialang/original/3X/e/4/e4ef04f5522c2677fbd3bfd3f80181716ff41ce7.png)

Now I’d like to **show the mean for each group** on the plot, like a scatter point. I see [this can be done](https://stackoverflow.com/a/74877951/17789881) for single boxplots, however for grouped ones I don’t know how to do it.

The best I could reach is grouping the DF so to get the mean for each group:

```julia
julia> df_grouped = groupby(df_temp, [:mnemonic, :forwarding]; sort = true);

julia> df_means = combine(df_grouped, :value => mean)
90×3 DataFrame
 Row │ mnemonic forwarding value_mean  
     │ String Bool Quantity…   
─────┼───────────────────────────────────
   1 │ adc false 0.0836535 W
   2 │ adc true 0.0863453 W
   3 │ add false 0.0839741 W
   4 │ add true 0.0828457 W
   5 │ and false 0.0834052 W
   6 │ and true 0.082732 W
   7 │ asr false 0.0861718 W
   8 │ asr true 0.0865888 W
   9 │ b false 0.170137 W
  10 │ bfc false 0.0876756 W
  11 │ bfi false 0.0854044 W
  12 │ bic false 0.0830427 W
  13 │ bl false 0.195068 W
  14 │ blx false 0.0943027 W
  ⋮ │ ⋮ ⋮ ⋮
  77 │ teq false 0.0844381 W
  78 │ tst false 0.0842573 W
  79 │ ubfx false 0.0855511 W
  80 │ ubfx true 0.0843686 W
  81 │ udiv false 0.0689451 W
  82 │ udiv true 0.073053 W
  83 │ umlal false 0.0848933 W
  84 │ umlal true 0.0915934 W
  85 │ umull false 0.083016 W
  86 │ umull true 0.0831577 W
  87 │ uxtb false 0.0744491 W
  88 │ uxtb true 0.0745435 W
  89 │ uxth false 0.0742926 W
  90 │ uxth true 0.0743984 W
                          62 rows omitted

```

**Is there a way I can do this?** Thanks.

---

<div class="post-metadata">

### Author: ![rafael.guerra](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/rafael.guerra/32/216610_2.png) [@rafael.guerra](https://discourse.julialang.org/u/rafael.guerra)
#### Post date: [August 19, 2023, 3:09pm UTC](https://discourse.julialang.org/t/how-can-i-show-the-mean-in-a-groupedboxplot/102980/2 "2023-08-19T15:09:44Z")

</div>

I don’t know of an automatic way, but if it helps, here’s a semi-automatic way.

> **CODE: add mean to groupedboxplot()**
>
> ```julia
> using StatsPlots, DataFrames
> 
> Random.seed!(123)
> mnemonics = ["add","and","lsl","mvn","rsb"]
> df = DataFrame(a=rand(Bool,200), b=rand(mnemonics,200), c=randn(200))
> 
> gdf = groupby(df, [:a,:b])
> df_means = combine(gdf, :c => mean)
> 
> p = @df df groupedboxplot(:b, :c, group=:a, label=["not active" "active"])
> 
> dic1 = Dict(xticks(p)[1][2] .=> xticks(p)[1][1])
> dic2 = Dict([false,true] .=> [-0.2, 0.2])
> x = getindex.(Ref(dic1), df_means.b) .+ getindex.(Ref(dic2), df_means.a)
> 
> scatter!(x, df_means.c_mean, c=:black, label = "mean")
> 
> ```

 ![StatsPlots_groupedboxplot_scatter_mean](https://global.discourse-cdn.com/julialang/original/3X/d/b/db7f564e02a2d62234d3ace385cd13e277c4c80d.png)

---

<div class="post-metadata">

### Author: ![bertulli](https://avatars.discourse-cdn.com/v4/letter/b/ecc23a/32.png) [@bertulli](https://discourse.julialang.org/u/bertulli)
#### Post date: [August 19, 2023, 9:54pm UTC](https://discourse.julialang.org/t/how-can-i-show-the-mean-in-a-groupedboxplot/102980/3 "2023-08-19T21:54:07Z")

</div>

Thanks, this works! It seems quite laborious though (it’s far over my Julia knowledge), and it’s a pity that Plots.jl and the recipe ecosystem (like StatsPlots.jl) are IMO so undocumented.  
Since in my code I also changed `bar_width`, I find a good solution to set

```julia
scatter_offset = bar_width / 4
dic2 = Dict([false,true] .=> [-scatter_offset, scatter_offset])

```

---

<div class="post-metadata">

### Author: ![rocco\_sprmnt21](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/rocco_sprmnt21/32/20127_2.png) [@rocco\_sprmnt21](https://discourse.julialang.org/u/rocco_sprmnt21)
#### Post date: [August 20, 2023, 9:40am UTC](https://discourse.julialang.org/t/how-can-i-show-the-mean-in-a-groupedboxplot/102980/4 "2023-08-20T09:40:08Z")

</div>

I don’t know if you find the following scheme simpler, which preorders the dataframes to make the positions of the scatter points correspond to the positions of the boxblocks in a “simpler” way

```julia
gdf = groupby(df, [:b,:a], sort=true)

df_means = combine(gdf, :c => mean)

p = @df sort(df,[:b,:a]) groupedboxplot(:b, :c, group=:a, label=["not active" "active"])

x=mapreduce(x->(x.+[-0.2,0.2]),vcat,xticks(p)[1][1])

scatter!(x, df_means.c_mean, c=:black, label = "mean")

```

---

<div class="post-metadata">

### Author: ![bertulli](https://avatars.discourse-cdn.com/v4/letter/b/ecc23a/32.png) [@bertulli](https://discourse.julialang.org/u/bertulli)
#### Post date: [August 20, 2023, 4:03pm UTC](https://discourse.julialang.org/t/how-can-i-show-the-mean-in-a-groupedboxplot/102980/5 "2023-08-20T16:03:16Z")

</div>

That is a possibility, thanks! It is a bit easier to move around if I don’t need a specific order for the bins.