AHL 4.12—Data collection, reliability and validity tests

$df=\text{categories}-1-\text{parameters estimated}$.

There are $8$ categories and $1$ parameter ($\lambda$) estimated from the data.

$df=8-1-1=6$.

The correct degrees of freedom is $6$.

$7$ comes from forgetting to subtract the estimated parameter $\lambda$, only applying the $-1$ for total probability.

$5$ results from mistakenly subtracting $2$ parameters, as if fitting a normal distribution instead of a Poisson.

$8$ ignores both the $-1$ constraint and the parameter subtraction entirely.

Only $df=6$ correctly applies both subtractions.

A double-barrelled question asks about two separate attributes within a single question.

"Easy to use" and "reliable" are two distinct attributes bundled into one question.

A respondent finding the app easy but unreliable cannot answer honestly with one response.

Option A is the double-barrelled question.

This asks about a single attribute (ease of use) using a defined scale - well constructed.

This asks about a single behaviour (frequency of use) with no ambiguity.

This is a leading question, not double-barrelled - it pushes toward a positive answer but only addresses one concept (quality/value combined loosely, but the defining flaw here is bias, not two separate measurable constructs).

Only option A explicitly combines two separate measurable constructs.