Data Analysis with Microsoft Excel: Updated for Office 2007

232 Fundamentals of Statistics

pressure or that polls predict a certain election result, you should ask, “And

what is the confi dence interval?”

If you want to generate a new set of random samples, you can click the

Generate Random Samples button in the Confi dence Intervals workbook.

Continue exploring the workbook until you understand the relationship

among the confi dence interval, the sample average, and the value of m. Close

the workbook when you’re fi nished. You do not have to save your changes.

Hypothesis Testing

Confi dence intervals are one way of performing statistical inference; another

way is hypothesis testing. In a hypothesis test, you formulate a theory about the

phenomenon you’re studying and examine whether that explanation is sup-

ported by the statistical evidence. In statistics, we formulate a hypothesis fi rst,

then collect data, and then perform a statistical test. The order is important. If

we formulate our hypothesis after collecting the data, we run the risk of hav-

ing a biased test, because our hypothesis might be designed to fi t the data. To

guard against a biased test, the hypothesis should be tested on a new set of data.

Figure 6-6 displays a classical approach to developing and testing a theory.

Observe phenomenon

Formulate a hypothesis

Collect data

Analyze data

Does analysis support

the hypothesis?

Report conclusions

Formulate a new hypothesis

no

yes

Figure 6-6
The steps in
developing
and testing a
hypothesis