What is a major threat to validity in a training outcomes study if baseline performance is not measured?

What is a major threat to validity in a training outcomes study if baseline performance is not measured?

• A. Baseline performance can confound results.
• B. Randomization ensures equivalence.
• C. A large sample size reduces threats.
• D. Lack of a control group.

Answer: (A)

Baseline performance refers to where participants start before the training. If you don’t measure it, differences in post-training outcomes may reflect who started higher or lower, not the effect of the training itself. This is confounding: the starting level is mixed with the training effect, making it impossible to tell what really caused the change. This threatens internal validity because you can’t confidently attribute observed improvements to the training. Even with random assignment, you’d still want baseline data to confirm groups were equivalent at the start; without it, you can’t verify that any differences after training aren’t just due to initial abilities. A large sample size helps with power and precision but doesn’t resolve this confounding. And while a control group is useful for causal inference, the specific issue here is the unmeasured starting point, which can mislead conclusions about training effectiveness.

Baseline performance refers to where participants start before the training. If you don’t measure it, differences in post-training outcomes may reflect who started higher or lower, not the effect of the training itself. This is confounding: the starting level is mixed with the training effect, making it impossible to tell what really caused the change. This threatens internal validity because you can’t confidently attribute observed improvements to the training.

Even with random assignment, you’d still want baseline data to confirm groups were equivalent at the start; without it, you can’t verify that any differences after training aren’t just due to initial abilities. A large sample size helps with power and precision but doesn’t resolve this confounding. And while a control group is useful for causal inference, the specific issue here is the unmeasured starting point, which can mislead conclusions about training effectiveness.