Skip to main content

07. Analyse Trial Results

Retrieve Output IDs​

To access trial results, first retrieve the available output IDs.

from jinko import JinkoClient

client = JinkoClient()
trial = client.get_trial("tr-abc123def456")

# Retrieving the output IDs
output_ids = trial.output_ids()
print("Available time series:", output_ids)

The response provides a list of available time series data points.

Retrieve Time Series Data​

Once you have identified the output IDs, you can request time series data. trial.results.timeseries(...) handles the download and decompression for you and returns a TabularDownload, which can be converted to a pandas dataframe with .to_dataframe() (requires pandas to be installed — it's an optional dependency of jinko-sdk).

TIME_SERIES_IDS = ["x"]

summary = trial.results.summary()
arms = summary["arms"]

timeseries_download = trial.results.timeseries(
    {timeseries_id: arms for timeseries_id in TIME_SERIES_IDS}
)
df_time_series = timeseries_download.to_dataframe()

Visualizing Time Series Data​

The resulting dataframe can be used directly for analysis and visualization.

import matplotlib.pyplot as plt

# Filter data for a single patient
unique_patient_ids = df_time_series["Patient Id"].unique()
patient_data = df_time_series[df_time_series["Patient Id"] == unique_patient_ids[0]]

# Plot each arm separately
for arm, group in patient_data.groupby("Arm"):
    plt.plot(group["Time"], group["Value"], marker="o", linestyle="-", label=arm)

# Customize the plot
plt.title("Time Series of Selected Output")
plt.xlabel("Time (seconds)")
plt.ylabel("Value")
plt.legend(title="Arm")
plt.show()

Summary​

In this module, you:

  • Retrieved the available output IDs for a trial.
  • Requested time series data as a pandas dataframe via trial.results.timeseries(...).to_dataframe().
  • Processed and visualized the results using Python and Matplotlib.

Next, you can explore more advanced analyses, such as aggregating results across patients with trial.results.scalars(...) and trial.results.scalars_per_population(...), or applying statistical models to the data.