Working with Your Own Data in Python

A language model can describe your data. It cannot tell you whether the pattern you are looking at is real or whether you are looking at ordinary variation. That is a different question and it needs a different tool. This course is for people who have understood what generative AI does and does not do, and who now need to work with their own figures directly instead of describing them to an assistant.ng from nothing, you will learn enough Python to load a dataset, select and sort and filter it, chart it, write your own functions, and then simulate: rather than trusting a formula, you will make the computer run the situation ten thousand times and watch what chance alone produces. By the end you will be able to say whether a result in your own data is surprising, and you will know which questions belong to an assistant and which belong to you. The material is based, with permission, on the University of California Berkeley Data 8 course, and has been delivered repeatedly to adult professionals.  

By the end of the training, participants will be able to:

• Write and run Python from scratch, with no prior programming experience

• Load a dataset and select, filter, sort and summarise it

• Produce histograms and charts that show what the data does rather than what you hoped

• Write your own functions and apply them across a whole dataset

• Simulate a situation many times to see what chance alone would produce

• Test whether a model of how something works is consistent with the data you observed

• Reason about samples: what a sample can and cannot tell you about the whole

• Recognise which questions a language model can answer and which require working with the data directly
  

Contents (8 Hours), ten units, each combining a short presentation of the concepts with hands-on lab work in a hosted Python notebook:

1. Getting started: notebooks, numbers, names, and reading an error message.

2. Tables: selecting, dropping, sorting and filtering a real dataset (Exercise).

3. Text and arrays, and loading your own data from a file (Exercise).

4. Table manipulation and visualisation: charts and histograms (Exercise).

5. Writing your own functions, and applying them across a whole table (Exercise).

6. Working through a real dataset end to end (Exercise).

7. Simulation: randomness, loops, and running a situation ten thousand times (Exercise).

8. Chance: probability worked out by simulation rather than by formula (Exercise).

9. Drawing random samples, and testing whether a model of how something works fits the data (Exercise).

10. Sampling variability and the normal distribution: how much a result moves by chance alone. Wrapup of the course.

Short presentations of each concept followed by lab work in a hosted Python notebook. Ten units across the day, with individual help during the labs, and exercises that can be completed afterwards at home.

Professionals who work with data in their organisation and have already understood the limits of generative AI tools: analysts, controllers, risk and compliance officers, fund and banking operations, HR and marketing analysts. Also suited to participants of Mastering Corporate AI Implementation who identified the data role as the one their organisation needs to fill. No programming experience required.

At the end of the training course, participants will be able to download a certificate of attendance issued by the House of Training from the learner portal.