13.11.2012 Views

Introduction to Categorical Data Analysis

Introduction to Categorical Data Analysis

Introduction to Categorical Data Analysis

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

2.6 EXACT INFERENCE FOR SMALL SAMPLES 47<br />

2, 3, 4). The observed table, three correct guesses of the four cups having milk added<br />

first, has probability<br />

P(3) =<br />

� �� �<br />

4 4<br />

3 1<br />

� �<br />

8<br />

4<br />

= [4!/(3!)(1!)][4!/(1!)(3!)]<br />

[8!/(4!)(4!)]<br />

= 16<br />

= 0.229<br />

70<br />

For Ha: θ>1, the only table that is more extreme consists of four correct guesses.<br />

It has n11 = n22 = 4 and n12 = n21 = 0, and a probability of<br />

� �� ��� �<br />

4 4 8<br />

P(4) =<br />

= 1/70 = 0.014<br />

4 0 4<br />

Table 2.9 summarizes the possible values of n11 and their probabilities.<br />

The P -value for Ha: θ>1 equals the right-tail probability that n11 is at least as<br />

large as observed; that is, P = P(3) + P(4) = 0.243. This is not much evidence<br />

against H0: independence. The experiment did not establish an association between<br />

the actual order of pouring and the guess, but it is difficult <strong>to</strong> show effects with such<br />

a small sample.<br />

For the potential n11 values, Table 2.9 shows P -values for Ha: θ>1. If the tea<br />

taster had guessed all cups correctly (i.e., n11 = 4), the observed result would have<br />

been the most extreme possible in the right tail of the hypergeometric distribution.<br />

Then, P = P(4) = 0.014, giving more reason <strong>to</strong> believe her claim.<br />

2.6.3 P -values and Conservatism for Actual P (Type I Error)<br />

The two-sided alternative Ha: θ �= 1 is the general alternative of statistical dependence,<br />

as in chi-squared tests. Its exact P -value is usually defined as the two-tailed<br />

sum of the probabilities of tables no more likely than the observed table. For Table 2.8,<br />

summing all probabilities that are no greater than the probability P(3) = 0.229 of<br />

Table 2.9. Hypergeometric Distribution for Tables with<br />

Margins of Table 2.8<br />

n11 Probability P -value X 2<br />

0 0.014 1.000 8.0<br />

1 0.229 0.986 2.0<br />

2 0.514 0.757 0.0<br />

3 0.229 0.243 2.0<br />

4 0.014 0.014 8.0<br />

Note: P -value refers <strong>to</strong> right-tail hypergeometric probability for one-sided<br />

alternative.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!