Putting Linear Trendline on a Plotly Subplot

Question

I wanted to know if there was an easier way I could put a linear regression line on a plotly subplot. The code I made below does not appear to be efficient and it makes it difficult to add annotations to the graph for the linear trendlines, which I want placed on the graph. Furthermore, it is hard to make axes and titles with this code.

I was wondering if there was a way I could create a go.Figure and somehow put it on the subplot. I have tried that, but plotly will only allow me to put the data from the figure on the subplot rather than the actual Figure, so I lose the title, axis, and trendline information. In addition, the trendline is hidden on the graphs because the scatterplot is overlaid on top of it. I tried changing how the data was displayed with data=(data[1],data[0]), but that did not work.

Basically, I want to know if there is a more efficient way of putting a trendline on the scatter plots than I pursued, so I can make it easier to set axes, set the graph size, create legends, etc, since it is difficult to work with what I coded .

sheets_dict=pd.ExcelFile('10.05.22_EMS172LabReport1.xlsx')
sheets_list=np.array(sheets_dict.sheet_names[2:])

fig=make_subplots(rows=7,cols=1)

i=0
for name in sheets_list:
    df=sheets_dict.parse(name)
    df.columns=df.columns.str.replace(' ','')
    df=df.drop(df.index[0])
    slope,y_int=np.polyfit(df.CURR1,df.VOLT1,1)
    LR="Linear Fit: {:,.3e}x + {:,.3e}".format(slope,y_int)
    rmse=np.sqrt(sum(slope*df.CURR1+y_int-df.VOLT1)**2)
    df['Best Fit']=slope*df.CURR1+y_int
    i+=1
    fig.add_trace(
        go.Scatter(name='Best Fit Line'+" ± {:,.3e}V".format(rmse),x=df['CURR1'],y=df['Best Fit'],
                   mode='lines',line_color='red',line_width=2),row=i, col=1)
    fig.add_trace(
        go.Scatter(name='Voltage',x=df['CURR1'],y=df['VOLT1'],mode='markers'),
        row=i, col=1)
#     fig.data = (fig.data[1],fig.data[0])

fig.show()

You have quite a lot of subplots – it would probably be more readable if you change the aspect ratio to be longer. For example, try `fig.update_layout(height=2000, width=800)`. Then you could try passing the trendline equation to the trace itself using `go.Scatter(name='Best Fit Line', ... text = [LR],...)`. if this helps, i'd be happy to write this up as an answer using some sample data — Derek O, Oct 11 '22 at 01:09
If you could write it up as an answer using sample data, that would be great. I was going to put my own data, but it is a giant Excel file and I couldn't do it justice by incorporating a sample of it. Thanks a lot! — dlnvss, Oct 12 '22 at 05:47

Whale Shark · Answer 1 · 2022-10-11T01:39:11.503

You can add titles and axes labels as follows:

import pandas as pd
import plotly.subplots as ps

fig=ps.make_subplots(rows=5,cols=1,subplot_titles=['Plot 1', 'Plot 2', 'Plot 3', 'Plot 4', 'Plot 5'])

fig.add_scatter(y=[2, 1, 3], row=1, col=1)
fig.add_scatter(y=[3, 1, 5], row=2, col=1)
fig.add_scatter(y=[2, 6, 3], row=3, col=1)
fig.add_scatter(y=[4, 0, 3], row=4, col=1)
fig.add_scatter(y=[3, 2, 3], row=5, col=1)
fig['layout']['xaxis']['title']='X-axis 1'
fig['layout']['xaxis2']['title']='X-axis 2'
fig['layout']['xaxis3']['title']='X-axis 3'
fig['layout']['xaxis4']['title']='X-axis 4'
fig['layout']['xaxis5']['title']='X-axis 5'
fig['layout']['yaxis']['title']='Y-axis 1'
fig['layout']['yaxis2']['title']='Y-axis 2'
fig['layout']['yaxis3']['title']='Y-axis 3'
fig['layout']['yaxis4']['title']='Y-axis 4'
fig['layout']['yaxis5']['title']='Y-axis 5'

fig.show()

The subplot_titles parameter in the make_subplots function is used to add plot titles. The fig['layout']['(x/y)axis(number)']['title'] is used to set the axes labels. Alternatively you can use:

fig.update_yaxes(title_text="yaxis 1 title", row=1, col=1)

or

fig.update_xaxes(title_text="xaxis 1 title", row=1, col=1)

To alter the plot sizes or spacing you can play around with the column_widths/row_heights or vertical_spacing/horizontal_spacing parameters of make_subplots: https://plotly.com/python-api-reference/plotly.subplots.html#subplots

As for the legend, there's no direct way of associating a legend with a subplot other than what you already have but the first comment in the following link shows a way of adding an annotation on the subplot that can act sort of like a legend: https://community.plotly.com/t/associating-subplots-legends-with-each-subplot-and-formatting-subplot-titles/33786

score 1 · Accepted Answer · answered Oct 11 '22 at 03:39

Trendlines are implemented in plotly.express with extensive functionality. See here. It is possible to create a subplot using that graph data, but I have created a subplot with a graph object to take advantage of your current code.

Since you did not provide specific data, I used the example data in ref. It is a data frame showing the rate of change in stock prices for several companies. It is in the form of a trend line added to it.

As for the graph, I have changed the height because a subplot requires height. The addition of axis labels for each subplot is specified in a matrix. If you need axis titles for all subplots, add them. Also, as a customization of the legend, we have grouped A group for the torrent lines and a group for the rate of change. As an example of the annotations, the slope values are set to 0 on the x-axis of each subplot and the y-axis is set to the position of the maximum value of each value.

import plotly.express as px
import plotly.graph_objects as go
import numpy as np

df = px.data.stocks()

df.head()
date    GOOG    AAPL    AMZN    FB  NFLX    MSFT
0   2018-01-01  1.000000    1.000000    1.000000    1.000000    1.000000    1.000000
1   2018-01-08  1.018172    1.011943    1.061881    0.959968    1.053526    1.015988
2   2018-01-15  1.032008    1.019771    1.053240    0.970243    1.049860    1.020524
3   2018-01-22  1.066783    0.980057    1.140676    1.016858    1.307681    1.066561
4   2018-01-29  1.008773    0.917143    1.163374    1.018357    1.273537    1.040708

from plotly.subplots import make_subplots

fig = make_subplots(rows=6,cols=1, subplot_titles=df.columns[1:].tolist())

for i,c in enumerate(df.columns[1:]):
    dff = df[[c]].copy()
    slope,y_int=np.polyfit(dff.index, dff[c], 1)
    LR="Linear Fit: {:,.3e}x + {:,.3e}".format(slope,y_int)
    rmse=np.sqrt(sum(slope*dff.index+y_int-df[c])**2)
    dff['Best Fit'] = slope*df.index+y_int
    fig.add_trace(go.Scatter(
        name='Best Fit Line'+" ± {:,.3e}V".format(rmse),
        x=dff.index,
        y=dff['Best Fit'],
        mode='lines',
        line_color='blue',
        line_width=2,
        legendgroup='group1',
        legendgrouptitle_text='Trendline'), row=i+1, col=1)
    fig.add_trace(go.Scatter(
        x=dff.index,
        y=dff[c],
        legendgroup='group2',
        legendgrouptitle_text='Rate of change',
        mode='markers+lines', name=c), row=i+1, col=1)
    fig.add_annotation(x=0.1,
                       y=dff[c].max(),
                       xref='x',
                       yref='y',
                       text='{:,.3e}'.format(rmse),
                       showarrow=False,
                       yshift=5, row=i+1, col=1)

fig.update_layout(autosize=True, height=800, title_text="Stock and Trendline")
fig.update_xaxes(title_text="index", row=6, col=1)
fig.update_yaxes(title_text="Rate of change", row=3, col=1)

fig.show()

Is there a way to put the trendline over the data? If I have a lot of data, it does not become possible to see the trendline. — dlnvss, Oct 12 '22 at 05:45
The order of the codes for the trendline and scatterplot should be scatterplot first, trendline second. — r-beginners, Oct 12 '22 at 06:56

Putting Linear Trendline on a Plotly Subplot

2 Answers2