10 New ways for you to Visualize your data.
Disclaimer: There might be visuals here that you already know of, don’t skip them, try to understand how you can make them look better or interpret them better.
Often we see Data being presented to us in reports everywhere in multiple forms of charts and graphs but most of them still use Bar Charts, Histograms and Pie Charts etc. which might be useful to show some amount of data. Data can be categorical or continuous but sometimes a Bar Chart or a Histogram is not enough to put forth our point. Their usefulness costs the reports it’s attention grabbing strength.
Just by putting in a little bit of work can make your reports look better with the use of some different visuals that look appealing to the audience it is presented to and that can be interpreted in a much more comfortable manner and are not difficult to prepare in any visualization tool. In this short article we’ll go over a set of 10 such visualizations that you may not have used in your reports and that can add to your reports the much needed touch of aesthetics and appeal.
Following is a list of the visuals, skip any visual that you already know of (considering the Disclaimer in the beginning of this article).
- Packed Bubbles
- Tree Maps
- Heat Maps
- Scatter Plots
- Box Plots
- Candle Sticks
- Chord Charts
- Contour Plots
- Sunbursts
- Polar Graphs
All the visuals ahead in this article are sourced/prepared using Tableau Public. (except Contour Plots)
Packed Bubbles:
Packed Bubbles are a go to when you have to highlight and show the proportions of your data and particularly depict where the majority of your data lies. It can be color coded according to any categories or intervals and can clearly depict data better than a Pie Chart can. They look better and can show significantly more number of variables than a Pie Chart.
They are suitable to concise your data (with many variables) into small packed bubbles that are appealing as well as easily interpretable.
Tree Maps:
A tree map can make categorical differentiation of variables as shown above. Here the world is divided into regions and the size of each rectangle shows the relative proportion of the specific country against the total population of the world. Though the name Tree Map isn’t synonymous to how a Tree Map actually looks, it still is easily interpretable as it gives out more details than a Packed bubble too (provided there are nested categories).
A tree map can show multiple categories within each categories clearly along with visible labels helping the audience understand the categorical hierarchy as well. Thus making it a good option for datasets with nested categories.
Heatmaps:
Heatmaps are by far the best method to see concentration of data over a wide range of variables easily. It is generally combined with a correlation matrix to see the relative correlation between variables easily and see which variables correlate with which one easily using the color saturation levels.
A darker shade generally represents a higher value in Heatmaps while a lighter shade represents a smaller value. It can also be done using multiple colors which follow a gradient. A labelled heatmap is similar to a Tree Map but used in terms of continuous variables and not usually for categorical variables. They can also be easily communicated with an audience and are not difficult to prepare.
Scatterplots:
Scatterplots are mainly used to identify relationships between variables and form clusters/segments in two continuous variables. They’re also used in fitting a regression line on the dataset to predict future values. Generally a scatterplot can show only two variables and can differentiate data on the basis of the number of clusters formed.
The points, as the name suggests, are directly plotted against the (X,Y) values and are analyzed to see if there’s a pattern in their placement. The markers can be opaque or partially transparent based on the need of the analysis. The markers can also change shapes into hexagons, triangles, squares, or ‘+’ sign as shown in the plot above. The density of the points also show the existence of a centroid and the identification of anomalies or outliers is very simple in a scatterplot. Hence, a lot of information can be gathered by using a simple scatterplot that’s easy to make in any tool.
Box Plots:
Box plots show the minimum value, the first quartile (25%), the second quartile (50%), the Median (M), the third quartile (75%), the fourth quartile (100%), the maximum value and any outliers (in a bottom to top order) As you see in the image above, there are outliers in the fourth box plot from the right and they exist outside the plot and are easy to identify.
A box plot communicates way more information than a normal histogram and can also show the concentration of data. It is always preferred over a simple histogram because of the amount of information it gives.
Also, as you can see, making multiple histograms for the data wouldn’t be easy and interpretable but making multiple box plots on multiple variables in the data is easy to understand and concise enough to fit in the size of a single histogram.
Candle Sticks:
Candle sticks are generally used in the Trading industry where the price of stocks are represented using these candle sticks. They include information about the highest price, the opening price, the closing price and the lowest price of the shares of a company over a period of time. They are similar to box plots but because of their monotonous color scheming, they cannot serve the purpose of box plots. Nevertheless, an interesting way to present data indeed.
If you’re into the Stock markets, try visualizing stocks in a candle stick plot of your own and see how it goes!
Chord Charts:
A chord chart is used to represent changes in variables or categories. If your categories are interrelated and have internal changes, a chord chart would be the way to go to visualize this in a simple way. Though it looks complex, upon observation of single categories, it is clear that a majority of the share of Samsung moves out and goes to other while it also constitutes of the highest share out of the total. The chord chart can be used if you want to keep track of debtors and creditors in finance or Incomes and expenditures in your daily life.
Showing the same data over a tree map or pie charts won’t reveal the underlying changes in the data as clearly depicted in a Chord Chart. It is easy to make using a Viz tool like Tableau and Power BI both.
Contour Plots:
Contour plots, though not easy to make on Tableau can depict information on a three dimensional (3D) setting.
Contour plots are generally used to identify ranges of a 3 dimensional dataset having a similar level or value. It is used in methods such as Gradient Descent of the Cost Function in machine learning models to arrive at a minimum point of the Cost. These plots also can be used against a Kernel Density Estimation plot to find a peak or fall in the range. As seen in the image above, there is a surface area depicted along with it’s color gradient using a Contour plot.
In case you have any ranges to visualize over 3D data, then try using a Contour plot!
Sunbursts:
A good alternative to the routine Pie Chart, a Sunburst plot works well with nested categorical variables and can show the proportion of a nested category against the entire dataset. It can be looked upon as the next level of a Donut chart. Sunburst charts can host more variables than a Pie Chart and can be labelled without seeming crammed up.
It is useful to also understand how data is distributed in the dataset. The color coding in this can be categorical or based on a gradient. This is one of the two alternatives to a Pie Chart, the next one is similar to a sunburst but can be used to explicitly show the differences in the sizes of a category against another.
Polar Graphs:
A polar graph is the next best alternative to a Pie Chart (at this point, you must be assuming I hate Pie charts, but that’s not so!). It is similar to a sunburst but the size proportions are visible clearer here compared to the first. Relative Multi-categorical proportions can also be differentiated and understood which wouldn’t be possible in a pie chart.
Also, in any of these graphs, as you can see in the image above, an annotation can be put in to bring the attention of an audience to a particular part of the plot (January in the image).
Note: Pie charts are the least recommended charts when data is presented in a report anywhere because it cannot support many variables and because it can be easily misinterpreted. From me to you, avoid them as much as possible. Instead take any of the available alternatives or a bar plot if nothing works.
While it is possible that these plots serve a particular need in a Viz, it is also possible that there are very similar plots to the ones that can vary over a small aspect and still make a big difference to the level of communication of the data to the viewers. A thing to note here is that, no matter how much you understand your data, Visualizations are to be made simple and it has to be ensured that your viewers perceive the data as closely as possible to your understanding.
This helps them have better intuitions about the data and take better decisions ahead if any are to be taken.
Most of the Visualizations around the world can be made using Tableau or Power BI. It is imminent in the field of analytics to be able to express your findings to your audience. This is possible using such Data Visualization tools that not just make your viz task easy but also gives your reports the appeal it needs to draw the attention of everyone.
Go ahead to Tableau Public’s website and start on your first beautiful visualization today!
For more info on tools that you can use, Refer to my previous article on The Data Science Starter Pack;
For more such articles, stay tuned, as I chart out the path on understanding data and coding and demystify other concepts related to Data Science and Coding. Please leave a review down in the comments. It was a long article, thank you very much for reading it all the way here! Great going!
