Unit 07: Analyzing Data and Visualizations

Data Analysis Process

A systematic approach to analyzing data, consisting of steps to collect, clean, visualize, and generate information.

Step 1: Collect or Choose Data

Gather the data needed for analysis.

Step 2: Clean/Filter

Remove errors or inconsistencies from the data. Focus on relevant data for analysis.

Step 3: Visualize and Find Patterns

Use graphs/charts to observe data for patterns.

Step 4: Generate New Information

Produce results based on observations.

Data

Information collected for analysis.

Metadata

Data about data including time of data collection, type of data, location of data collection, method of collection, and collector of the data.

Bar Charts

Visualizations that can be vertical or horizontal, showing frequency analysis where taller/longer bars indicate more frequent values.

Insights from Bar Charts

Identify most and least common values, range, and presence of values.

Pie Charts

Visualizations that represent percentages of unique values in a dataset.

Insights from Pie Charts

Identify highest/lowest percentages and compare values.

Histograms

Displays frequency of values within ranges and reads similarly to bar charts.

Insights from Histograms

Identify most and least common ranges.

Scatterplots

Compares two data columns to find relationships, which can be direct, inverse, or none.

Insights from Scatterplots

Identify relationships and trends; make predictions.

Correlation

Indicates similarities and apparent patterns between data sets.

Causation

Implies one event causes another.

Important Note on Correlation and Causation

Correlation does not equal causation.

Big Data

Data gathered through data mining and web scraping, solving problems like efficiency in business and disease identification.

Open Data

Freely available data with minimal restrictions, sourced from open data repositories.

Crowdsourced Data

Data collected by ordinary people for decision-making.

Machine Learning

Involves algorithms that analyze data and adapt, used in daily tasks and AI.

Limitations and Bias in Machine Learning

Algorithms may reflect human biases if the input data is not diverse.

Example of Bias

Twitter's cropping algorithm favored certain demographics due to biased training data.

Ways to Mitigate Bias

Diversify training data by including underrepresented groups.

Simulation

A model of real-world situations/events useful for hypothesis testing when real experimentation is impractical or risky.

Usage of Simulations

Help abstract complex processes and provide insights that cannot be easily realized in real life.