Exam DP-100 topic 5 question 7 discussion

"A two-tailed test is appropriate if the estimated value is greater or less than a certain range of values, for example, whether a test taker may score above or below a specific range of scores." https://en.wikipedia.org/wiki/One-_and_two-tailed_tests

While the question does not state whether the datasets are paired, it actually seems more likely that the data is paired. Here’s why: "Estimated values in one set of data may be more than or less than reference values in the other set of data": This wording suggests a comparison between estimated values and reference values. Often, estimated values are paired with reference values to evaluate accuracy or difference. For example, you might be comparing predicted values with observed values, which is a classic paired scenario. "Constant Type I error as a function of correlation": Maintaining a constant Type I error rate as a function of correlation typically implies that the correlation between two sets of data points (e.g., estimated vs. reference) is being taken into account. This is a key aspect of a paired t-test, where the test accounts for the natural pairing and potential correlation within the pairs. That said, D is correct answer. Probably. Maybe.

If we assume that the two data sets are uncorrelated, then choose the unpaired t-test. If the two data sets are related (for example, two measurements from the same set of samples), choose a paired t-test. The question does not clearly state whether the data sets are paired, but based on common data comparison scenarios, we can infer that they are paired.

i think the given answer is correct

Option C is the correct answer: Paired t-test with a one-tail option.

I guess two tails due to two data set are extremely negatively correlated??? Not sure whether that is the ask from the question though, cannot find any reference from azureml Purely statistically speaking, I agree with one tail paired test

Answer is D Paired t test with two test. But I highly doubt this question is still being asked

Is this question still being asked?

A distribution that has a constant Type I error as a function of the correlation. = Paired

It should be Unpaired as Each pair of scores is independent of every other pair. And , one tail, we are looking for only inequality(more or less)

This picture is taken from Wikipedia: https://en.wikipedia.org/wiki/Student's_t-test#Unpaired_and_paired_two-sample_t-tests

Why paired?