How do you learn and improve your data visualization skills in Python?

To improve your data visualization skills in Python, focus on mastering key libraries like Matplotlib, Seaborn, and Plotly. Start by understanding the basics of each library and their unique strengths. Practice creating various types of plots, from simple line charts to complex heatmaps and interactive visualizations. Experiment with customizing colors, styles, and layouts to enhance readability. Regularly work with real datasets to apply your skills in practical scenarios. Stay updated with the latest features and best practices by following tutorials, online courses, and data visualization communities. Seek feedback on your visualizations and continuously refine your techniques to create more effective and impactful data representations.

Hay varias opciones que puedes considerar. Una opción es explorar las bibliotecas de visualización de datos de Python, como Matplotlib, Seaborn y Plotly. Estas bibliotecas proporcionan una amplia gama de herramientas para crear visualizaciones de datos interactivas y atractivas. Otra opción es seguir cursos en línea o tutoriales que se centren en la visualización de datos en Python. Hay muchos recursos en línea que pueden ayudarte a aprender las mejores prácticas y conceptos de visualización de datos en Python. Además, puedes practicar la visualización de datos en Python utilizando conjuntos de datos de muestra y experimentando con diferentes tipos de gráficos y técnicas de visualización.

Matplotlib is a very versatile option for quickly producing plots in Python. By modifying the rc parameters (rcParams), you can customise the look and style to enhance the quality of your graphs and make them suitable for journals or better fit the design style of your publication. I personally use import matplotlib as mpl mpl.rc('font', family=u"serif", style="italic", variant="normal", weight="normal", size=24) to improve the fonts on all my graphs and other pre-sets to standardise appearance across all my graphs. As I work with numpy and scipy reguarly, it's a convenient library for extracting outputs from those libraries and creating publication quality graphs.

Python libraries that are in nearly all of my scripts/programs are; Matplotlib, Pandas, NumPy, and BioPython. These libraries have robust documentation and tutorials, where I never really felt lost while learning how to properly and efficiently utilize them. Nearly every single publication image generated using Python will typically include one or more of these libraries. I highly recommend you start with these, after developing a fundamental understanding of base Python and its syntax.

The Software Carpentry lessons are amazing, free resources for learning programming and data visualization in general, and python specifically. https://software-carpentry.org/lessons/

Clarity is often about removing information, rather than adding as much as possible. A standardised colour palette (that adheres to guidelines for visual impairments) makes your graphs recognisable and clear to look at but being mindful of line widths, distances between elements (data points, curve fits, error bars, text, etc) and ensuring the image isn't cluttered is an important, albeit sometimes forgotten, art of a good graph. Once you have your base information required to make your point (data, fit, uncertainties), add the minimal amount of necessary information (usually just the axes labels) and let the caption and the accompanying text provide any clarification that can't be conveyed both visually and simply.

The visualization must narrate a story. A key rule I learned from my PhD supervisor was that a chart/figure must stand independently so that almost no words are required to explain it. When analyzing data during research, I tend to use plotly as its plots are dynamic allowing me to zoom in/out and interact with the data for deeper analysis. When producing final figures, I combine Matplotlib and Seaborn for visual appeal. I like to keep things simple, least white space, and a colorblind safe color scheme Choosing an intuitive color scheme is also important for eg using different shades of the same color can effectively represent data points that share similarities.

For myself, I started by working through various "cook books" on GitHub, and toying around with my own data sets to build a "comfortable" foundation. To expand my Python skills to program/software development for biological data, I began cloning repositories of tools I routinely used from GitHub and backward engineer the scripts and structure, to gain a better understanding of how Python is used in the biological sciences. My university also offered free day-long workshops on Python and sponsored certificate training in AI/ML training with Python by NVIDIA. Always leverage your local academic institutions, you will be surprised at the free training available to you.

Subplots are an excellent way to compare different datasets or variables side by side. In a financial analysis project where I compared stock prices from multiple companies, I experimented with Matplotlib’s subplot function to display price trends, trading volume, and moving averages on the same figure. This method enabled me to observe correlations and divergences between variables at a glance. Subplots facilitate direct comparisons between datasets, making it easier to identify patterns and anomalies in a consolidated visual framework.

Whilst experimenting with visualising your own data, you may find it useful to do so within Jupyter notebooks. The notebook environment enables you to compare plots inplace within the document, and avoid cluttering your file system with various test figures. If you are particularly happy with a visualization you have achieved, the notebook environment further enables you to provide documentation describing how you rendered the figure, supplemented with a real example you have produced with your code.

One key aspect of getting familiar and more intuitive with the plotting and visualization is playing with the plots. I randomly have selected CSV files and tried different plots on them. It gave me an intuition upon reading csv’s , which plot might be more accurate in terms of enhanced explainability. Hence get your hands dirty and be street smart in plots.