Math 143 C/E, Spring 2001
IPS Reading Questions
Chapter 7, Section 1 (pp. 502-507; up to The one-sample
t test)
While the information in Section 6.1 is accurate, carrying out the confidence interval procedure that is described there requires that you know the standard deviation of the population. Now the whole setting assumes that you don't know (but want to know) something about the population mean. If you don't know the population mean, isn't it even less likely for you to know its standard deviation? While knowledge of the population standard deviation is probably impossible, you can calculate an approximation to it, namely s, the standard deviation for the sample. Using it (instead of s) you can also compute an approximate spread (referred to as the standard error of the sample mean; see p. 504) for the sampling distribution of the mean. Nevertheless, use of standard error in place of s introduces more complications than you might expect, the most important of which is that the sampling distribution is no longer a normal distribution. Rather, it is a t distribution.
The number of degrees of freedom is one less than the sample size. For samples of size 20, you would use the t distribution with df = 19 (i.e., 19 degrees of freedom).
After closing up the previous applet, click here and here to start up applets in two more windows. One applet draws the student t distribution, allowing you to set the degrees of freedom and move sliders around to highlight (in yellow) the region whose area you want to know. The other applet is similar, but uses the standard normal distribution. Set the degrees of freedom to 1 for the student t curve, and highlight the region between 0 and 1 (these values located along the horizontal axis). Compare the percentage of area between 0 and 1 for the two curves. Keep raising the number of degrees of freedom for the t curve until the amount of area is pretty much the same. What was the least number of degrees of freedom required? If you consider the area between 0 and 2 at this number of degrees of freedom, is it still the same?
The value of 34.1% that the standard normal curve gives isn't seen under the student t curve until you reach 144 degrees of freedom. Even at this level, the areas under the two curves between 0 and 2 do not match exactly.
We learned in Chapter 6 (when we assumed that s was known) that the margin of error for population means is m = z* s. When a t distribution is required (that is, when s is not known), the critical z* value is replaced with the critical t* value and the standard deviation for the population s is replaced with that (s) for the sample.
df | C | t* |
17 | 90% | |
28 | 99.5% | |
80 | 98% |
Answers: 1.740, 3.047, 2.374