Shift Plot
Shift plots (also known as difference plots or before-after plots) are used to visualize the differences between two related measurements. They are particularly useful for paired data analysis, before-after comparisons, and treatment effect visualization.
Note: The figures on this page are generated by running examples/shiftplot_examples.py and examples/shiftplot_extra_examples.py using identical parameters.
Features
Paired data visualization: Shows relationships between two related measurements
Difference highlighting: Clearly displays the magnitude and direction of changes
Statistical annotations: Optional statistical tests for paired comparisons
Custom styling: Flexible color and marker options
Publication-ready: Clean, professional appearance
Basic Usage
from ggpubpy import plot_shift, load_iris
import matplotlib.pyplot as plt
# Use Iris data: Setosa vs Versicolor sepal length (basic)
iris = load_iris()
x = iris[iris["species"] == "setosa"]["sepal_length"].values
y = iris[iris["species"] == "versicolor"]["sepal_length"].values
fig = plot_shift(x, y)
plt.show()
Function Parameters
plot_shift()
Parameters:
x(array-like): First measurement (e.g., before treatment)y(array-like): Second measurement (e.g., after treatment)x_label(str, optional): Label for x-axisy_label(str, optional): Label for y-axistitle(str, optional): Plot titlefigsize(tuple): Figure size (default: (8, 6))alpha(float): Transparency for points (default: 0.7)color(str): Color for points and lines (default: ‘#2E86AB’)line_color(str): Color for connecting lines (default: ‘#A23B72’)
Returns:
Figure: matplotlib figure. Access axes viafig.axes[0].
Examples
Shift Plot with Quantile Connectors (Main only)
from ggpubpy import plot_shift, load_iris
import matplotlib.pyplot as plt
import numpy as np
# Use Iris data: Setosa vs Versicolor sepal length
iris = load_iris()
x = iris[iris["species"] == "setosa"]["sepal_length"].values
y = iris[iris["species"] == "versicolor"]["sepal_length"].values
# Main plot with quantile connectors (no bottom subplot)
fig = plot_shift(
x,
y,
paired=False,
n_boot=1000,
percentiles=[10, 50, 90],
confidence=0.95,
violin=True,
show_quantiles=True,
show_quantile_diff=False,
x_label="Setosa",
y_label="Versicolor",
title="Iris: Setosa vs Versicolor Shift Plot",
subtitle="Main plot with quantile connectors",
color="#27AE60",
line_color="#2C3E50",
alpha=0.8,
)
plt.show()
Shift Plot with Quantile Differences Subplot
from ggpubpy import plot_shift, load_iris
import matplotlib.pyplot as plt
import numpy as np
# Iris data: Setosa vs Versicolor
iris = load_iris()
x = iris[iris["species"] == "setosa"]["sepal_length"].values
y = iris[iris["species"] == "versicolor"]["sepal_length"].values
# Main plot + quantile differences subplot
fig = plot_shift(
x,
y,
paired=False,
n_boot=1000,
percentiles=[10, 50, 90],
confidence=0.95,
violin=True,
show_quantiles=True,
show_quantile_diff=True,
x_label="Setosa",
y_label="Versicolor",
title="Iris: Setosa vs Versicolor Shift Plot",
subtitle="With quantile differences subplot",
color="#27AE60",
line_color="#2C3E50",
alpha=0.8,
)
plt.show()
When to Use Shift Plots
Shift plots are particularly useful for:
Before-after studies: Compare measurements before and after an intervention
Paired data analysis: Visualize relationships between related measurements
Treatment effects: Show individual responses to treatments
Quality control: Monitor changes in processes or products
Longitudinal studies: Track changes over time in the same subjects
Interpretation
Key Elements
Points: Each point represents a pair of measurements
Lines: Connect each pair, showing the direction and magnitude of change
Diagonal line: Reference line where x = y (no change)
Position relative to diagonal:
Above diagonal: y > x (increase/improvement)
Below diagonal: y < x (decrease/decline)
On diagonal: y = x (no change)
Statistical Information
The plot includes:
Paired t-test: Tests if the mean difference is significantly different from zero
Effect size: Cohen’s d for the paired difference
Confidence interval: 95% CI for the mean difference
Tips
Sample size: Works best with moderate to large sample sizes (n > 20)
Outliers: Be aware of extreme values that might skew the interpretation
Color choices: Use contrasting colors for points and lines for better visibility
Transparency: Adjust alpha to handle overlapping points
Reference line: The diagonal line helps interpret the direction of changes
Statistical tests: Use the built-in statistical annotations to quantify effects
Integration
from ggpubpy import plot_shift, plot_boxplot_with_stats, load_iris
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
# Iris dataset: Setosa vs Versicolor
iris = load_iris()
x = iris[iris["species"] == "setosa"]["sepal_length"].values
y = iris[iris["species"] == "versicolor"]["sepal_length"].values
# Shift plot for group comparison
fig_shift = plot_shift(x, y, x_label="Setosa", y_label="Versicolor", title="Integration: Shift Plot (Iris)")
# Box plot for distribution comparison
df = pd.DataFrame({
'Group': ['Setosa'] * len(x) + ['Versicolor'] * len(y),
'Value': np.concatenate([x, y])
})
fig_box, ax_box = plot_boxplot_with_stats(df, x="Group", y="Value", x_label="Species", y_label="Sepal Length (cm)")
plt.show()
Advanced Usage
from ggpubpy import plot_shift
from scipy import stats
import matplotlib.pyplot as plt
import numpy as np
# Create paired data
np.random.seed(42)
n = 30
before = np.random.normal(10, 2, n)
after = before + np.random.normal(1, 1.5, n)
# Create plot
fig = plot_shift(before, after, x_label="Before", y_label="After", title="Advanced: Shift with custom stat", subtitle="Wilcoxon signed-rank")
# Add custom statistical information
ax = fig.axes[0]
wilcoxon_stat, wilcoxon_p = stats.wilcoxon(before, after)
ax.text(0.02, 0.98, f'Wilcoxon signed-rank test:\np = {wilcoxon_p:.3f}',
transform=ax.transAxes, fontsize=10, verticalalignment='top',
bbox=dict(boxstyle="round,pad=0.3", facecolor="lightgreen", alpha=0.7))
plt.show()
