Data Science Applications to Astronomy

Week 11: Communicating Data Science:

Reports & Dashboards

Reports

• You (or your team) performed a detailed analysis.

• A report shares your conclusions and recommendations with others.

Who are you reporting to?

• Your team members (e.g., regular group meeting)

• A wider collaboration (e.g., internal white paper, annual collaboration meeting)

• Experts beyond your collaboration (e.g., journal article, scientific meeting)

• Funding agency (e.g., progress/final grant report)

• General public (e.g., press release, funding agency)

• Planning committee (e.g., white paper for decadal survey)

What are you aiming to accomplish from the report?

• Stimulate suggestions for how to solve current roadblock

• Provide documentation for team members to build on your results

• Help identify future opportunities for connections to other researchers/projects

• Get critical feedback to identify weak points of analysis

• Provide documentation to justify continued/increased support for project

Prepare report to advance specific goals

• Use the knowledge gained from performing a detailed analysis positions

• Narrative & choice of figures (in report body) directly support report's goals

Typical Scientific Report Structure

1. Summary of Conclusions/Recommendations

   • Abstract

   • Executive Summary

   • Opening paragraphs "above the fold"

2. Context:

   • Why did you do this work?

     • Problem or unmet need

     • Previous studies left unanswered question

   • What other information is needed to understand your approach and results?

   • Why did you choose this approach, as opposed to alternatives?

3. Describe Input Data

   • Where did data come from?

   • How was data collected?

   • What is known (or unknown) about quality of data?

   • What concerns about data should be kept in mind?

4. Describe Data Analysis & Outputs

   • How were data analyzed?

   • What are the outputs of the analysis?

   • Sometimes includes limited straight-forward results

5. Interpret Data

   • Explain rationale for each conclusion/recommendation

6. Conclusions/Recommendations

   • Concise summary

7. What comes next?

   • Motivate what you want to do next

   • Inspire others to contribute

8. Appendices

   • Supporting Data, Tables, Figures, Codes

   • More details about Analysis Methods & Implementation Details

   • Describe checks that were performed and passed

Pros of Typical Written Scientific Report

• Provides details needed to understand, evaluate and/or build on results

• Great for referring back to long after report was prepared

• Sense of accomplishment from finishing a project

Cons of Typical Written Scientific Report

• Can take significant time to prepare

• Can take significant time to understand

• Easy to overlook important elements "buried" in a long report

• Readers may skim/skip over parts

• Often "dry"

Oral Reports/Presentations

• Summarizes key elements of written report

• Refers to written report for details (e.g., data, methodology, analysis, complete list of assumptions/caveats)

  • Should still highlight most important assumptions/caveats

• Best presentations tell a story

  • Easier to maintain audience's attention

  • More likely to be remembered

Common Presentation Structures

• Solve a Problem/Need

  • Problem... How Found the Solution... Solution

  • Problem... Solution... Reasoning

• The Big Idea

  • One one hand... On the other hand... Call to Action...

  • What is... What could be... Call to Action...

• Hero's Journey (hopefully abridged from Campbell version)

Common Presentation Elements

• Polished opening to capture's audience's attention/imagination

• Asking audience a question

• Dramatic pause / use of empty space

• Call to Action

Dashboard

Goal: Efficiently communicate what can be learned from data

• Support people in making future decisions

• Won't represent the best-possible analysis

• Automates performing common tasks & analysis

• Needs to be robust & capable of handling corner cases gracefully

Example Dashboard Structure

• Select which data to display

• Summary/simplified version of data

• Temporal evolution of data and/or summary

• Comparison dataset or model

  • Select which comparison to make

  • Visualization of comparison data/model

  • Summary of comparison data/model

  • Temporal evolution of comparison data/model (on top of data?)

• Opportunities to dig deeper to better understand numbers

  • Minimally processed data avaliable

  • Drilldown to help users find data of interest/concern

Examples of Dashboards at Penn State

• Penn State Undergraduate Admissions Dashboard

• Penn State Graduation & Retension

• Penn State Community Survey

• University Park Energy Usage

• Penn State Data Center Utilization

Examples of Regional Dashboards

• Census Data

• PA Wastewater Survelience System

Pros of Dashboard

• Enables people to gain insights efficiently

• Accelerate science

• Reduce errors by automating routine steps

• Allows team members to focus on other steps

• Can integrates expertise from multiple team members

Cons of Dashboard

• Takes significant time to make pipeline robust

• Takes significant time to design effective visualizations

• May not be able to refer back to dashboard's state at previous time

• Takes time/knowledge to keep running smoothly

• Often want to update dashboard (repeatedly) as needs change

• Risks oversimplifying complex cases

• Can contribute to leaders becoming overconfident

Key Differences between a Report & a Dashboard

Dashboards aim to

• Reduce cognitive load of viewer...

  • but provide all information necessary

• Anticipate questions that viewers will have...

  • but questions change with time/new data

• Be flexible enough to deal with whatever data comes in...

  • but incoming data changes with time

Q&A

Accessibility

Projects

For presentation, the audience is your peers in the class.

For dashboard, your README can state your intended audience. Some possibilities:

• Another member of your research group (potentially most helpful)

• A researcher you've never met

• An observer/telescope operator

• Another member of this class (reasonable default option)

PlutoPlotly

using PlutoPlotly
layout = Layout(  title="Mt Bruno Elevation",  autosize=false,  width=500,  height=500,  margin=attr(l=65, r=50, b=65, t=90) ) 
plot(surface(z=z_data), layout)

Plotly Gallery

Data Disagregation & Drilldown

Discovery Method: AllTransitRadial VelocityMicrolensingImagingTransit Timing VariationsAstrometryPulsar TimingOrbital Brightness ModulationEclipse Timing VariationsPulsation Timing VariationsDisk Kinematics

Show Drilldown Plots

Misc

Deep Learning Frameworks

• Python-based

  • Tensorflow

  • PyTorch

• JAX

  • Equinox

• Julia-based

  • Flux.jl

  • Lux.jl

  • TensorFlow.jl

  • Torch.jl

  • SimpleChains.jl

Collaborating

Asynchronous

• Write separate files/functions/modules

• Maintain independent repositories

• Merge changes via git

• Create branches for new features, so main branch is always usable

Synchronous

• Like asynchronous, but ask questions as you go

• Pair Coding: Driver & Navigator

• Debugging: Explainer & Audience

• Beware of using shared filesystem

Tools

• Merging updates: Git

• Sharing screen: Zoom, Teams, Virtual Desktop

• Collaborative coding: VS Code/VS Codium, Repl.it, Google Colab

Setup/Helper Code