Interacting with tables in ChatGPT
When you upload a file, ChatGPT will automatically create an interactive table view that allows you to scroll through your data and view all of your rows and columns.
Once the file is uploaded, you can follow up with questions pertaining to the dataset. The prompts do not need to specify specific operations - natural language commands like “analyze” or “compare” are sufficient for producing results from ChatGPT.
You can also create tables natively in ChatGPT by requesting that the output is generated as a table.
To get a better look at your data, you can expand the table by clicking on the two arrows on the top-right corner of the table:
In the table view, you can select a specific row or column and create a prompt to obtain insights about the data highlighted. For example, here we have selected a column and asked for the item that appears most frequently.
Multiple rows or columns can be selected by pressing and holding the Command key on Mac or the Ctrl key on Windows and clicking on the rows or columns. Multiple cells can also be selected by clicking on one cells and dragging the mouse to cover the intended area.
After selecting multiple items, you can ask ChatGPT to calculate a value or perform an action on the values of the selected. For instance, you can highlight a set of cells and ask ChatGPT to compute the average value.
Editing and creating tables with ChatGPT
You can upload and edit your existing tables by prompting ChatGPT to make updates. For example, you can prompt ChatGPT to update a table with a column containing the average values.
You can prompt ChatGPT to make specific changes by highlighting the columns, rows, or cells that you want the updates based on. Here, we have highlighted two of the columns and requested a new column to the table that contains their sums.
You can download the table generated by ChatGPT by clicking on the download button on the top-right corner of the table. Please note that the downloaded file will be in a CSV format.
Visualizing your data with ChatGPT
After uploading a file, you can prompt ChatGPT to produce a static chart. You can allow ChatGPT to determine the ideal chart type for the dataset, or specify one of our supported chart types in your prompt: line graph, bar chart, pie chart, histograms, scatter plot, box plots (Box-and-Whisker Plots), heat maps, area charts, radar charts, treemaps, bubble charts, and waterfall charts.
Please note that only bar, pie, scatter, and line charts are currently interactive in most cases.
If no chart type is specified, ChatGPT will determine the ideal chart type to output.
On the top-right corner of the chart, you can download or expand the size of the chart. By default, downloaded charts are in a PNG format.
You can also make edits to chart colors or toggle its interactivity on or off. When changing colors, you can select one of our default colors or input the hex code of a color.
Common types of analysis
ChatGPT is trained to perform a variety of data analysis tasks. Some common tasks include:
Anomaly detection and mitigation
When making decisions with data, it’s important to ensure that your source data is as accurate as possible. ChatGPT knows how to identify data which might be missing or incorrect. Common issues ChatGPT can identify and repair include:
Missing values
Outlier values
Duplicate rows
Incorrect data types
Start your analysis with a prompt like this: Check this data for common issues.
Once ChatGPT has identified common issues, you can ask it to fix those issues. Depending on the issues encountered, ChatGPT may offer multiple options for you to choose from. If you’re uncertain about the implications of these choices, try asking ChatGPT for more information.
Aggregation & integration
ChatGPT can aggregate large amounts of structured data to help you make sense of information. Some aggregations ChatGPT can perform include:
Sums
Averages (median, mean, mode)
Minimum and maximum values
Counts of distinct values
Standard deviation
You can expand a table and select one or more numerical columns, and then use a prompt like this: Calculate the median and standard deviation for this data.
ChatGPT can also merge multiple datasets together based on shared identifiers.
Let’s say you upload two spreadsheets, one containing customers and one containing purchases. Purchase records are associated with customer records via a customer_id property. ChatGPT knows how to integrate both files into a single dataset so that it can answer questions like "What is the total of all purchases made by customers with a gold plan?”
ChatGPT automatically merges datasets for you when you ask a question where it is required.
Advanced statistical analysis
ChatGPT understands how to perform a wide variety of statistical analyses, and is able to select appropriate techniques based on your requirements. Some types of analyses ChatGPT can perform include:
Comparative statistics: This involves comparing different groups or variables to understand their differences or similarities. Techniques include t-tests, ANOVA (Analysis of Variance), and MANOVA (Multivariate Analysis of Variance).
Correlation and regression Analysis: These methods assess the relationship between variables. Correlation analysis measures the strength and direction of the relationship, while regression analysis models the relationship to predict outcomes.
Time Series analysis: This type of analysis examines data points collected over time to identify trends, cycles, and seasonal effects. Methods include ARIMA (Autoregressive Integrated Moving Average) and Seasonal Decomposition.
You don’t need to be a data scientist to use these techniques! If you’re not sure which technique is most appropriate for your data, try telling ChatGPT what you want to understand, and ask it to recommend the best analysis technique. If you’re not sure how to interpret the output of the analysis, ask ChatGPT to explain it to you. An effective prompt can be: Is there anything notable or unusual about this analysis?