OpenEDG PCAD flashcards

32 cards. Read the term and the definition here. The deck at the top of this page is optional.

raise_for_status()

A requests.Response method that raises requests.HTTPError when the status code is 4xx or 5xx, converting silent bad responses into exceptions you can handle.

API pagination

Splitting a large API result into pages using offset/limit, page numbers, or cursor tokens; the client loops until no next token or empty page is returned.

Exponential backoff

A retry strategy where the wait time grows multiplicatively per attempt, often with random jitter, used to handle rate limits and transient server errors gracefully.

pd.read_sql

A pandas function that executes a SQL query through a DBAPI or SQLAlchemy connection and returns the result set as a DataFrame with column names as labels.

Many-to-many merge blowup

When joining on a key that is not unique on both sides, every matching pair produces a row, inflating row counts; validate key uniqueness before merging.

errors='coerce'

An argument to pd.to_numeric and pd.to_datetime that replaces unparseable values with NaN instead of raising, allowing you to detect and handle bad records.

IQR outlier rule

Flags a value as an outlier when it falls below Q1 minus 1.5 times the interquartile range or above Q3 plus 1.5 times the IQR.

drop_duplicates

A pandas method removing duplicate rows, optionally restricted to a subset of columns; the keep parameter selects which occurrence (first, last, or none) is retained.

Referential integrity check

A validation confirming every foreign key value in a child table exists in the referenced parent table, catching orphaned or mistyped records before analysis.

loc versus iloc

loc selects by row and column labels and includes the endpoint; iloc selects by integer position and excludes the endpoint, matching Python slicing.

NumPy broadcasting

Shape alignment from the rightmost dimension, where dimensions of size one or missing dimensions are stretched to allow elementwise operations without copying data.

pivot_table

A pandas method producing a spreadsheet-style summary that groups by index and column keys and aggregates values with a function such as mean or sum.

axis=0 versus axis=1

In NumPy and pandas reductions, axis=0 collapses rows to give a per-column result, while axis=1 collapses columns to give a per-row result.

SQL logical execution order

FROM and JOIN, then WHERE, GROUP BY, HAVING, SELECT, DISTINCT, ORDER BY, LIMIT; WHERE filters rows before grouping and HAVING filters aggregated groups.

LEFT JOIN

A join returning every row from the left table, with matching right-table columns filled in or NULL where no match exists, preserving unmatched left records.

COUNT(*) versus COUNT(col)

COUNT(*) counts all rows including those with NULLs, while COUNT(col) counts only rows where that column is not NULL, so results can differ.

ROW_NUMBER window function

Assigns consecutive integers to rows within a partition according to an ORDER BY clause, with no ties and no gaps, useful for deduplication and top-n per group.

Parameterized query

A SQL statement with placeholders whose values are supplied separately by the driver, preventing SQL injection because user input is never parsed as SQL code.

Standard error of the mean

The standard deviation of the sampling distribution of the sample mean, computed as s divided by the square root of n, shrinking as sample size grows.

p-value

The probability of observing data at least as extreme as the sample, assuming the null hypothesis is true; compared with alpha to decide rejection.

Type I and Type II errors

Type I is rejecting a true null hypothesis, a false positive with probability alpha; Type II is failing to reject a false null, a false negative with probability beta.

Pearson versus Spearman correlation

Pearson measures the strength of a linear relationship between raw values; Spearman applies Pearson to ranks and captures any monotonic relationship, resisting outliers.

R-squared

The proportion of variance in the dependent variable explained by the model, computed as one minus the ratio of residual sum of squares to total sum of squares.

RMSE

The square root of mean squared error, expressing typical prediction error in the same units as the target while penalizing large errors more heavily than MAE.

Data leakage

When information from the test set or target influences training, such as fitting a scaler on all data, producing optimistic evaluation scores that fail in production.

L1 versus L2 regularization

Lasso adds the sum of absolute coefficients as a penalty and can zero coefficients out for selection; Ridge adds squared coefficients and shrinks them smoothly.

Precision and recall

Precision is true positives divided by predicted positives, measuring correctness of positive calls; recall is true positives divided by actual positives, measuring coverage.

Matplotlib object-oriented API

Creating fig, ax = plt.subplots() and calling methods on the axes object, which allows composing multiple plots and explicit control over each figure element.

Seaborn axes-level functions

Functions like scatterplot and boxplot that draw on a single Matplotlib Axes and accept an ax argument, unlike figure-level functions that create their own grid.

Chart type selection

Line for time trends, bar for category comparison, histogram for one distribution, scatter for two-variable relationships, and box plot for spread across groups.

super().__init__()

A call inside a subclass constructor that runs the parent class initialization, ensuring inherited attributes are set before the subclass adds its own state.

try/except/else/finally

A block where except handles matching exceptions, else runs when no exception occurred, and finally always runs, typically for cleanup such as closing files.