Hi,
I would like to create a BoxWhisker plot, however the dataset I’m working with is big and I’ve calculated the min, max, Q1, Q3, median and mean beforehand.
But in the examples I only see how to plot the BoxWhiskers by giving it the full dataset for which holoviews then calculates the stats.
Is there a way to create the BoxWhisker by just giving the 5 values for the positions of the whiskers and boxes?
Cheers!
Hacked something together, see below. But still interested to know how other people handle these cases where the samples are just very large and can’t be passed directly?
import holoviews as hv
import pandas as pd
hv.extension("bokeh")
df = pd.DataFrame(
{
"x": ["model 1", "model2"],
# NOTE whiskers
"y1": [-4.5, -2.6],
"y5": [14.5, 11.4],
# NOTE boxes
"y2": [2, 4],
"y4": [5, 8],
# NOTE markers
"y3": [2.7, 5.1],
}
)
z1 = df.assign(errneg=df["y3"] - df["y1"], errpos=df["y5"] - df["y3"])
z2 = df.assign(up=df["y4"] - df["y2"]).melt(
id_vars=["x"], value_vars=["y2", "up"], var_name="pos"
)
whiskers = hv.ErrorBars(z1, kdims=["x"], vdims=["y3", "errneg", "errpos"])
boxes = hv.Bars(z2, kdims=["x", "pos"]).opts(
cmap=["rgba(0,0,0,0)", "red"],
line_width=0,
stacked=True,
multi_level=True,
show_legend=False,
)
markers = hv.Points(df, kdims=["x", "y3"], vdims=["y1", "y2", "y4", "y5"]).opts(
size=75,
marker="dash",
tools=["hover"],
color="k")
X = hv.Overlay([whiskers, boxes, markers])