Okay, so you finished your data analysis and found that there were statistically significant differences. You then ran a post-hoc test and found exactly where these differences were. Now what? Graph time!
Choosing the right graph or chart for you data can be challenging. You can’t just plot a bunch of points and call it a graph. Your graph needs to display your findings clearly and effectively. When I was creating figures and tables for my thesis, I often had to revise them multiple times until I found the right visualization.
This guide will help you select the appropriate figure and troubleshoot problems that may arise when using Julius to create them.
Types of Data Visualizations
-
Bar Chart: Used for comparing values of different categories or groups.
Example: monthly sales performance of different products -
Line Chart: Ideal for showing trends over time or continuous data.
Example: Temperature variation over a week in city, with each point specifying the temperature at a specific time.
-
Pie Chart: Useful for displaying parts of a whole, though less popular in scientific articles.
Example: Distribution of a company’s quarterly expenses. -
Histogram: Effective for displaying the distribution of continuous data. Useful in exploratory analysis to examine skew, normality and kurtosis.
Example: Distribution of ages in a population sample. -
Scatterplot: Used for visualizing the relationship between two continuous variables.
Example: Relationship between study hours and exam scores for students, with study hours on the x-axis and exam scores on the y-axis. -
Box Plots (my favourite!): Helpful for displaying the distribution of datasets and highlighting outliers.
Example: Distribution of salaries in a company; showing minimum, maximum, median, and quartiles of the dataset. -
Heatmap: Displays patterns and relationships in large datasets.
Example: Correlation matrix of stock prices for portfolio of assets. Each cell in the map indicates a correlation coefficient between two of the stocks, with the warmer colours (red) representing stronger positive correlations and cooler colours (blue) representing strong negative correlations. -
Polar Plots (also a personal favourite): Uses polar coordinates to represent data, with radial and angular axis.
Example: Mapping animal movement patterns to see directional preferences.
Troubleshooting Graphs on Julius
Question 1a: I want Julius to create a simple bar graph of the monthly sales performance on different products.
Prompt 1a: Can you create a simple bar graph that displays the monthly sales performance of different products?
Correction: This graph is not what I wanted. Although it’s clean, I want a simple bar graph, not a stacked one. Let’s try again, being more specific:
Prompt 1b: Using the data from this chart, can you create a simple bar graph where each bar represents a product, showing the monthly sales performance across one year?
This is closer to what I wanted to display but still not perfect. Let’s refine the prompt:
Prompt 1c: Can you create a grouped bar chart illustrating the mean monthly sales performance of different products (Product A, Product B, Product C, and Product D) over a year, with four bars for each month?
This is the bar graph is closer to what I envisioned, but it is still messy. The bar graph from Prompt 1b would probably be better for showing annual mean sales comparisons. However, I want to see how sales change over time, so let’s try another visualization.
Prompt 2a: Can you create a line graph showing the trend of total sales over the year?
Julius created a graph showing the overall trend of sales over the year, but I wanted to see the trend for all four products. Let’s try again:
Almost done! I just want to make each graph a different colour and remove the headings and gridlines for a cleaner look.
Prompt 2b: Can you create a line graph for each product please?
Almost done! I just want to make each graph a different colour and remove the headings and gridlines to create a crisper look.
Prompt 2c: Can you make Product A blue, Product B orange, Product C green and Product D red please, and remove the headings and gride lines for each graph please?
For the guide, I combined these instructions into one prompt, but typically you should separate them.
Now we just need a legend!
Final Note: Include standard error bars or standard deviations on your graphs to show the variability of your dataset. Since I used a fictional dataset, I did not have the additional information to include these. However, your dataset should have them, and I encourage you to place them on any graph you make.
Happy graphing!
Keywords: AI, GPT 4, Claude 3, Data Analysis, Data Science, AI, Statistics, Statistical Analysis, Data Visualization, Graph Type, Charts
