Plotly Surface Plot with Categorical Data

Question

Trying to understand exactly how the Surface plot inputs work for plotly. Every example I've seen has all values be numeric, and the desired input is for x or y to be a category.

In a test example, I have 3 states that have a 'count' value for each week. I want to see the progression of each state's 'count' on a week by week basis. Data is initially structured like so

import pandas as p
import plotly.graph_objects as go

# dummy categorical data
labels = ['state','week','count']
test_data = [
['Washington', 1, 5],
['Washington', 2, 3],
['Washington', 3, 1],
['Washington', 4, 0],
['California', 1, 2],
['California', 2, 4],
['California', 3, 6],
['California', 4, 8],
['New York', 1, 0],
['New York', 2, 5],
['New York', 3, 5],
['New York', 4, 15]]
dataFrame = p.DataFrame(test_data, columns=labels)

where it gets tricky afterwards is I'm not sure how this needs to be structured so when viewing the plot you have

• an X axis containing each week
• a Y axis containing each state
• a Z axis showing the count for each state, on a week to week basis

---

I've tried a lot of different methods at this point, and I can tell it understands the max value for Z because the colorscale bar goes up to the max value in the dataframe but I can't figure out what I need to do so each state is properly displayed as its own tick, each week is displayed on the other axis, and the z corresponds directly with the week:state 'count' value from the dataframe. None of the attempted methods have been quite right with results so different it wouldn't add value to show them here.

Does anyone have a practical working example with real categorical data, or know the way to structure the above dataframe so the surface plot shows count by state over time?

---

A more specific way to ask it would be: how do you take a dataframe where 3 columns should be your X, Y, and Z and have them display as such in a Surface plot? between X and Y, one of them needs to be a string representing a category, and the other needs to be able to be a datetime or int.

Answer 1

Ah, it takes the same inputs as a heatmap. The X and Y are more so the label values while your z data is what's going to contain all the plotted values. I was able to put it together as so using a previous example

# grab x/y labels, create storage for z data
states = dataFrame['state'].unique()
states.sort()
weeks = dataFrame['week'].unique()
weeks.sort()
z_data = []

# extract z data using x,y coordinates within the dataframe
for week in weeks:
  row = []
  for state in states:
    val = dataFrame[
      (dataFrame['state'] == state) &
      (dataFrame['week'] == week)]['count'].values[0]
    row.append(val)
  z_data.append(row)


# create and show the Surface plot
trace = go.Surface(x=states, y=weeks, z=z_data)
data = [trace]
fig = go.Figure(data=data)
fig.show()

---