CQP Foundation Level

Quantitative Skills and Data Analytics flashcards

Read the question, say the answer to yourself, then turn the card to check. The first 10 of 50 cards are free; the rest come with the subject.

Card 1 of 50

Jump to a card

Complete Flashcard Reference

Review every card in this deck. Open any card to reveal its answer. The first 10 are free; the other 40 come with the subject.

What question does each of the four types of analytics answer?

Descriptive: what happened? Diagnostic: why did it happen? Predictive: what is likely to happen? Prescriptive: what should we do? Each builds on the one below it, and most organisations fail at the base of that stack rather than at the top.

Name the four levels of measurement and what sets each apart.

Nominal: categories with no order (defect type). Ordinal: ordered categories with unequal intervals (minor, major, critical). Interval: equal intervals, arbitrary zero (degrees Celsius). Ratio: equal intervals and a true zero (fill volume in ml). The level decides which statistics are permissible.

How should an analyst treat an outlier?

Investigate it, never delete it automatically. If it is an error, correct or remove it with the reason recorded. If it is a real, unusual observation, it is often the most informative point in the dataset: in quality work it is frequently the defect you were looking for.

What is a p-value, and how is it used to decide a hypothesis test?

The probability of a result at least as extreme as the one observed, if the null hypothesis were true. If p is less than alpha (conventionally 0.05), reject H0. Failing to reject H0 is not evidence that H0 is true: a small sample may simply lack the power to detect a real difference.

What is the difference between a primary key and a foreign key?

Primary key: a column, or combination of columns, that uniquely identifies each row. It cannot be null and cannot repeat. Foreign key: a column whose values must exist as a primary key in another table. It enforces referential integrity, so a production record cannot reference a machine that does not exist.

How are Cp and Cpk calculated, and what does each measure?

Cp = (USL - LSL) / (6s): spread only, whether the process could fit the tolerance if perfectly centred. Cpk = min[(USL - x-bar)/(3s), (x-bar - LSL)/(3s)]: spread and centring, whether it does. Cp is always greater than or equal to Cpk; the gap is the cost of being off centre.

Which chart suits each kind of question?

Comparison: bar chart. Trend over time: line chart. Composition: stacked bar or treemap. Distribution: histogram or box plot. Relationship: scatter plot. Choose the chart from the question, not from preference.

What is overfitting, and how is it detected?

A model that fits its training data so closely that it learns the noise as well as the relationship, and so predicts new data badly. Detect it by evaluating on held-back test data the model has never seen: a large gap between training and test performance (say 96% against 61%) is the signature.

Why is database access control a quality control and not only an IT one?

Segregation of duties: the person who records production cannot also authorise release. Record integrity: revoking DELETE on inspections enforces 'records are never altered' technically. Attribution: individual logins make the audit trail evidence of who did what. Personal data: the Data Protection Act, 2019 requires access to be limited to those with a legitimate purpose.

What must the executive summary of an analytics report contain?

The question, the answer and the recommendation, in under one page. It names the number and the action: which machine, from what to what, why, and what to do by when, with owner and cost. Write it last, from the completed analysis, so it can stand alone.

How do data, information and knowledge differ?