Analyze a sensor CSV with ChatGPT
For graduate researchers. Upload six fictional CO₂ readings, check the missing value, compare room means, and turn the result into a two-sentence lab update.
Read the full transcript
0:00 · Start with a research question
You are preparing a lab meeting update. You have three observation times in two rooms. Open the practice file below the player and use ChatGPT to calculate a comparison you can reproduce. Keep your spreadsheet open alongside the chat. By the end you will have a checked result, a note for your supervisor, and a prompt you can adapt to another experiment.
0:30 · Load the practice file
Start a new chat. Use the attachment control beside the message box to select room readings dot C S V. Wait until its name appears in the composer. If attachments are unavailable, open the practice file in a text editor, copy its contents, and paste it in the chat. The exercise works with either input.
1:00 · Check what the tool received
Send this prompt before calculating anything. Ask for column names, the number of data rows, and the location of empty cells. Compare the response with your file. The headers are room, time, and carbon dioxide in parts per million. Count six data rows. Room B at ten has no reading. An empty cell is not a measured zero.
1:30 · Give precise instructions
Ask for a separate mean for each room. Say to exclude blank readings. Use the field name exactly as it appears in the file so the tool uses the correct column. Request a table with valid count, sum, and mean. Those intermediate values let you catch a mistaken denominator quickly. Run the prompt, then pause to inspect your own results.
2:00 · Verify room A
Check the first row yourself. Six hundred plus seven hundred twenty plus eight hundred forty equals two thousand one hundred sixty. Divide by three to get seven hundred twenty. In a blank spreadsheet cell type equals average, open parenthesis, C two through C four, close parenthesis. If your result differs, compare the individual input cells with the practice file.
2:30 · Catch a denominator error
Check room B next. Eight hundred fifty plus nine hundred fifty is eighteen hundred. Divide by two valid readings. Your spreadsheet formula uses C five through C seven. If the tool reports six hundred, ask it to list every value used and its denominator. Then request a corrected table. Do not repair the lab note until the calculation is right.
3:00 · Make a comparison chart
Ask ChatGPT to draw a bar chart of the room means. Start the vertical axis at zero, label it mean carbon dioxide in parts per million, and include the valid counts. Compare each bar with your verified table. A label or unit can change the meaning of a chart. If your tool cannot draw it, use the checked table to create the chart in your spreadsheet.
3:30 · Write the lab update
Request two sentences for a lab meeting. Specify that the note must summarize these observations only, include valid counts and units, and mention the missing reading. Read the result. Nine hundred minus seven hundred twenty gives one hundred eighty. The file does not contain occupancy or ventilation settings. Remove any claim that explains the difference using either one.
4:00 · Try a sensitivity check
Try this follow-up in the same chat. Compare room means using only timestamps with a valid reading in both rooms. Ask which timestamp was dropped and why. Before you look at the answer, remove the ten o’clock observation from room A as well. Two observations remain per room. Predict whether the difference changes in this example.
4:30 · Save a reproducible result
At matched times, the means happen to stay the same. The counts change. This agreement is specific to the example. With different middle readings the two approaches could produce different conclusions. Save the original file, the prompt, the checked table, and your final two sentences together. For another department’s dataset, replace the grouping field, measurement, and units before running the same workflow.
Try it yourself
Download the practice CSV and open a new chat in ChatGPT. Attach the file or paste its contents. Send these prompts in order.
List the column names, number of data rows, and location of blank readings.
Group by room. Exclude blank co2_ppm readings. Show the valid count, sum, and mean in ppm for each room.
Plot the room means as a bar chart with a zero baseline. Label the y-axis Mean CO₂ (ppm) and annotate each bar with its valid count.
Draft two sentences for a lab meeting. Include counts, units, the difference between means, and the missing reading. Limit conclusions to these observations.
Repeat the comparison using only times with valid readings in both rooms. Identify the excluded time and report the new counts and means.
Check your results
A: 3 readings, mean 720 ppm. B: 2 readings, mean 900 ppm. At matched times (09:00 and 11:00), both counts are 2 and the means stay the same.