What a learner can do afterwards
- Chooses between comparing a mean with a known value and comparing two means, and runs the right test
- Compares two spreads with an F-test and says what a significant result implies about the methods
- States a null hypothesis, quotes the confidence level and interprets the outcome in chemical terms
- Applies an outlier test and explains why discarding a result needs a stated rule
1 · Read
Two results that differ may differ only because measurements scatter. A significance test states how likely that scatter is to explain the gap, which turns an argument about numbers into a decision with a stated risk. Start from the null hypothesis, the claim of no real difference, and quote your confidence level: 95 percent confidence leaves 5 percent, or 0.05, as the bar.
Javier and Ercilia each asked 100 shoppers about the trip to the mall, and both found the same average, 6.0 miles. But Javier's answers spread by about 4.0 miles while Ercilia's spread by about 7.0 miles. Trust the smaller spread: its average sits on steadier ground. A test that compares two spreads, such as an F-test, tells you whether that gap in scatter beats random noise, and a significant result means the precisions really differ, so you must not pool the two sets.
Pick the test that fits the shape of your data. One batch of readings against a certified value, like a balance checked against 100.0 mg, calls for a one-sample test. Morning and evening readings of the same flask come in pairs, so subtract each pair and test the single column of differences. Readings from different batches in two labs are independent, so compare the two means directly. A large P-value, like 0.40 against a 0.05 bar, means you fail to reject the null, which is never a proof that it is true.
Never throw out a result just because it looks odd. Discarding needs a rule you stated before testing, applied through a proper outlier test. And when a small P-value like 0.03 beats your 0.05 bar, call the result significant, then say in plain chemical terms what it does and does not prove.
Name the null, run the test that fits your data, and read the P-value as a decision with a stated risk.
2 · Watch
Take it off screen
Where it sits
8 questions wait behind this lesson, each with its answer explained. Every answer feeds the sky: stars light as they are learned, and dim when it is time to come back.