Your starting point
Prerequisites: Use the ideas and vocabulary from lesson 1; revisit its worked examples if needed.
Definition: Categorical data identifies groups; numerical data measures or counts quantities. Nominal categories have no inherent order. Ordinal categories have an order but not necessarily equal gaps. Discrete quantities are counts; continuous quantities are measured along a scale.
Meaning comes before appearance
A value made of digits is not automatically a quantity. A postal code, telephone number or product identifier is a label. Adding two product codes has no useful interpretation. A count such as number of orders is numerical because addition and differences have meaning. Ask what an operation would mean in the real situation rather than trusting a spreadsheet’s automatic formatting.
Respect the scale
Low, medium and high satisfaction are ordered categories. The jump from low to medium is not known to be the same size as the jump from medium to high. Coding them 1, 2 and 3 does not prove equal intervals. Report category counts or proportions unless you can justify stronger assumptions. For a measured duration, by contrast, a difference of five minutes has a consistent interpretation when collection is comparable.
Inspect mixed and missing values
An otherwise numeric column may contain “unknown,” a currency symbol or an accidental note. The workbench reports missing and invalid values separately. It does not guess that a blank equals zero. Type inference is a starting suggestion: inspect several records, the data dictionary and edge cases. User correction changes how a field is treated, but cannot repair an inaccurate source measurement. Preserve identifiers as text when leading zeros matter.
Two worked examples
Inventory identifiers
The inventory sample contains SKU 0001 and SKU 0002. These are nominal labels, even though every character is a digit. Their mean, 1.5, would not identify an average product. Stock values 5 and 8 are discrete counts: their sum of 13 units is meaningful if they refer to distinct inventory items.
Survey and timing
A survey records satisfaction as low, medium or high and waiting time as minutes. Satisfaction is ordinal; waiting time is numerical. A bar chart of satisfaction counts respects the categories. A histogram of waiting times shows a numerical distribution. A blank waiting time means no recorded duration; a zero means an observed duration of zero.
Where this becomes useful
A support manager combines ticket channel, priority and response duration. Channel is nominal, priority is ordered, and duration is measured. Treating every column as a number would obscure rather than simplify the analysis.
Make an attempt
Use the Data Type Classifier. Classify each field, explain why its operation is meaningful, and check the authored feedback. Then load Inventory and inspect the inferred SKU and stock types. Override one type and notice that the source strings are unchanged.
Misconception check
Claim: Assigning numbers to labels makes them numerical measurements.
Inspect the correction
Codes can encode categories without changing their measurement scale. Document what each code means.
Independent application
Create a five-field dictionary for a hypothetical booking table. Include a nominal field, an ordinal field, a count, a measured quantity and an identifier. For each, state one useful analysis and one operation that would be misleading.
Completion records practice and self-review. It is not certification of statistical competence.
