LUME #1

First collection for learn about how construction LUME

Flashcards:Educational Data Analysis Fundamentals

Educational Data Analysis Fundamentals

Educational Data Analysis FundamentalsPage 1 of 2

Front

Back

SQL GROUP BY Clause

A SQL clause used to aggregate and collapse multiple database rows into summary rows based on shared values in specified columns.

Groups identical values in specified columns so aggregate functions like COUNT, SUM, or AVG can operate on each group.
Think of it as creating 'buckets' for data, such as grouping student records by classroom or school year.
Essential in educational management to calculate aggregate metrics, like average attendance per school tier.
Common mistake: Selecting non-aggregated columns that are not included in the GROUP BY clause.

SQL HAVING Clause

A SQL clause used to filter aggregated group results after the GROUP BY clause has been processed.

Filters summary groups based on aggregate function calculations like COUNT() > 10 or AVG(gpa) < 7.5.
Remember: WHERE filters individual records before grouping, while HAVING filters groups after aggregation.
Used to isolate specific at-risk groups, such as classrooms where average attendance falls below a safety threshold.
Common mistake: Trying to use the WHERE clause instead of HAVING to filter on aggregate functions.

SQL Execution Order (WHERE vs HAVING)

The fixed logical sequence SQL follows, where WHERE filters raw rows before grouping and HAVING filters aggregate groups after.

Order of processing: FROM -> WHERE -> GROUP BY -> HAVING -> SELECT.
Mnemonic: 'We Have Groups' (WHERE comes before HAVING, separated by GROUP BY).
Crucial for writing accurate queries, ensuring inactive records are removed before calculating school group statistics.
Common mistake: Attempting to filter row-level attributes in HAVING or group aggregates in WHERE.

INNER JOIN

A SQL join operation that combines records from two tables only when there is a matching key in both tables.

Returns the intersection of two tables, dropping any unmatched rows from either side.
Associate with an 'exclusive overlap' where both dataset conditions must be present.
Useful for analyzing complete student records where both demographic and exam data are present.
Common mistake: Using INNER JOIN on educational data and accidentally excluding vulnerable students who lack records.

LEFT JOIN

A SQL join operation that retains all records from the left table and populates unmatched columns from the right table with NULL.

Guarantees that every row from the primary (left) dataset remains in the output.
Think of it as a 'safety net' join that preserves baseline student rosters regardless of secondary table matches.
Vital for spotting dropout risk because missing right-table data appears as explicit NULL values.
Common mistake: Placing the main student list on the right side of a LEFT JOIN instead of the left.

Educational Data Analysis FundamentalsPage 2 of 2

Front

Back

Pandas isna().sum()

A Pandas expression that calculates the total number of missing or NaN values within each column of a DataFrame.

Generates a boolean mask via .isna() (True=1, False=0) and sums the True values for every column.
Memory aid: 'Spot and Count' empty spaces in your data grid before running computations.
Helps educational researchers audit datasets to determine missing attendance logs or unrecorded grades.
Common mistake: Confusing the output count of missing values with the count of valid, present entries.

Pandas dropna() vs fillna()

Pandas methods for managing missing data, where dropna removes rows containing NaN values and fillna replaces NaNs with specified values.

dropna() deletes rows/columns with missing data; fillna() substitutes missing slots with constants or statistics like mean/median.
Association: dropna is 'erase', fillna is 'replace'.
Deciding between them determines whether at-risk students are removed from analysis or assigned baseline proxy values.
Common mistake: Carelessly using dropna() on key educational datasets, which unintentionally erases vulnerable students.

Label Encoding vs One-Hot Encoding

Techniques for converting text categories to numbers, where Label Encoding uses ordered integers and One-Hot Encoding creates binary columns.

Label Encoding transforms categories to 0, 1, 2 (preserves rank); One-Hot Encoding creates separate 0/1 indicator columns (no rank).
Use Label Encoding for ordinal data (Low, Medium, High risk) and One-Hot Encoding for nominal data (School Name, Region).
Ensures machine learning algorithms process student metadata without introducing false mathematical hierarchies.
Common mistake: Using One-Hot Encoding on ordered categories, which destroys natural ranking information.