Chapter 18: Inferential Statistics Answers to Review Questions 18.1.

What is the difference between a statistic and a parameter A statistic is a numerical characteristic of a sample, and a parameter is a numerical characteristic of a population. 18.!. What is the s"mbol for the population mean The symbol is the Greek letter mu (i.e., ). 18.#. What is the s"mbol for the population correlation coefficient The symbol is the Greek letter rho (i.e., ). 18.$. What is the definition of a samplin% distribution The sampling distribution is the theoretical probability distribution of the values of a statistic that results hen all possible random samples of a particular si!e are dra n from a population. 18.&. 'ow does the idea of repeated samplin% relate to the concept of a samplin% distribution "epeated sampling involves dra ing many or all possible samples from a population. 18.(. Which of the two t"pes of estimation do "ou li)e the most* and wh" This is an opinion #uestion. $oint estimation is nice because it provides an e%act point estimate of the population value. &t provides you ith the single best guess of the value of the population parameter. &nterval estimation is nice because it allo s you to make statements of confidence that an interval ill include the true population value. 18.+. What are the advanta%es of usin% interval estimation rather than point estimation The problem ith using a point estimate is that although it is the single best guess you can make about the value of a population parameter, it is also usually rong. Take a look at the sampling distribution of the mean on page '() and note that in that case if you ould have guessed *+,,,,, as the correct value (and this -A. the correct value in this case) you ould be rong most of the time. A ma/or advantage of using interval estimation is that you provide a range of values ith a kno n probability of capturing the population parameter (e.g., if you obtain from .$.. a 0+1 confidence interval you can claim to have 0+1 confidence that it ill include the true population parameter. An interval estimate (i.e., confidence intervals) also helps one to not be so confident that the population value is e%actly e#ual to the single point estimate.

That is, it makes us more careful in ho e interpret our data and helps keep us in proper perspective. Actually, perhaps the best thing of all to do is to provide both the point estimate and the interval estimate. 2or e%ample, our best estimate of the population mean is the value *34,(', (the point estimate) and our 0+1 confidence interval is *3,,053.65 to *3',3((.40. 7y the ay, note that the bigger your sample si!e, the more narro the confidence interval ill be. &f you ant narro (i.e., very precise) confidence intervals, then remember to include a lot of participants in your research study.

18.8 What is a null h"pothesis A null hypothesis is a statement about a population parameter. &t usually predicts no difference or no relationship in the population. The null hypothesis is the 8status #uo,9 the 8nothing ne ,9 or the 8business as usual9 hypothesis. &t is the hypothesis that is directly tested in hypothesis testing. 18.,. -o whom is the researcher similar to in h"pothesis testin%: the defense attorne" or the prosecutin% attorne" Wh" The researcher is similar to the prosecuting attorney is the sense that the researcher brings the null hypothesis 8to trial9 hen she believes there is probability strong evidence against the null. :ust as the prosecutor usually believes that the person on trial is not innocent, the researcher usually believes that the null hypothesis is not true. &n the court system the /ury must assume (by la ) that the person is innocent until the evidence clearly calls this assumption into #uestion; analogously, in hypothesis testing the researcher must assume (in order to use hypothesis testing) that the null hypothesis is true until the evidence calls this assumption into #uestion. 18.1.. What is the difference between a probabilit" value and the si%nificance level 7asically in hypothesis testing the goal is to see if the probability value is less than or e#ual to the significance level (i.e., is p < alpha). The probability value (also called the p=value) is the probability of the observed result found in your research study of occurring (or an even more e%treme result occurring), under the assumption that the null hypothesis is true (i.e., if the null ere true). >ote that the probability value is the most important concept in hypothesis testing. ?ake sure you kno hat it means and hat it does not mean@ &n hypothesis testing, the researcher assumes that the null hypothesis is true and then sees ho often the observed finding ould occur if this assumption ere true (i.e., the researcher determines the p value). The significance level (also called the alpha level) is the cutoff value the researcher selects and then uses to decide hen to re/ect the null hypothesis.

?ost researchers select the significance or alpha level of .,+ to use in their research; hence, they re/ect the null hypothesis hen the p value ( hich is obtained from the computer printout) is less than or e#ual to .,+. The key idea of hypothesis testing it that you re/ect the null hypothesis hen the p value is less than or e#ual to the significance level of.,+.

18.11. Wh" do educational researchers usuall" use ..& as their si%nificance level &t has become part of the statistical hypothesis testing culture. &t is a longstanding convention. &t reflects a concern over making type & errors (i.e., anting to avoid the situation here you re/ect the null hen it is true, that is, anting to avoid 8false positive9 errors). &f you set the significance level at .,+, then you ill only re/ect a true null hypothesis +1 or the time (i.e., you ill only make a type & error +1 of the time) in the long run. 18.1!. State the two decision ma)in% rules of h"pothesis testin%. "ule oneA &f the p value is less than or e#ual to the significance level then re/ect the null hypothesis and conclude that the research finding is statistically significant. "ule t oA &f the p value is greater than the significance level then you 8fail to re/ect9 the null hypothesis and conclude that the finding is not statistically significant. 18.1#. /o the followin% statements sound li)e t"pical null or alternative h"potheses 0a1 -he coin is fair. 0b1 -here is no difference between male and female incomes in the population. 0c1 -here is no correlation in the population. 0d1 -he patient is not sic) 0i.e.* is well1. 0e1 -he defendant is innocent. All of these sound like null alternative hypotheses (i.e., the 8nothing ne 9 or 8status #uo9 hypothesis). -e usually assume that a coin is fair in games of chance; hen testing the difference bet een male and female incomes in hypothesis testing e assume the null of no difference; hen testing the statistical significance of a correlation coefficient using hypothesis testing, e assume that the correlation in the population is !ero; in medical testing e assume the person does not have the illness until the medical tests suggest other ise; and in our system of /urisprudence e assume that a defendant is innocent until the evidence strongly suggests other ise. 18.1$. What is a -"pe I error What is a -"pe II error 'ow can "ou minimi2e the ris) of both of these t"pes of errors &n hypothesis testing there are t o possible errors e can makeA Type & and Type && errors. A Type & error occurs hen your re/ect a true null hypothesis (remember that hen the null hypothesis is true you hope to retain it); it also is called a 8false positive.9

A Type && error occurs hen you fail to re/ect a false null hypothesis (remember that hen the null hypothesis is false you hope to re/ect it); it also is called a 8false negative.9 The best ay to allo yourself to set a lo alpha level (i.e., to have a small chance of making a Type & error) and to have a good chance of re/ecting the null hen it is false (i.e., to have a small chance of making a Type && error) is to increase the "our sample si2e. The key in hypothesis testing is to use a large sample in your research study rather than a small sample@ &f you do re/ect your null hypothesis, then it is also essential that you determine hether the si!e of the relationship is practically significant (see the ne%t #uestion).

18.1&. If a findin% is statisticall" si%nificant* wh" is it also important to consider practical si%nificance -hen your finding is statistically significant all you kno is that your result ould be unlikely if the null hypothesis ere true and that you therefore have decided to re/ect your null hypothesis and to go ith your alternative hypothesis. Bnfortunately, this does not tell you anything about ho big of an effect is present or ho important the effect ould be for practical purposes. ThatCs hy once you determine that a finding is statistically significant you ne%t use one of the effect si!e indicators to tell you ho strong the relationship. Think about this effect si!e and the nature of your variables (e.g., is the &D easily manipulated in the real orldE -ill the amount of change relative to the costs in bringing this about be reasonableE). Fnce you consider these additional issues beyond statistical significance, you ill be ready to make a decision about the practical significance of your study results. 18.1(. 'ow do "ou write the null and alternative h"potheses for each of the followin%: 0a1 -he t test for independent samples* 0b1 3ne4wa" anal"sis of variance* 0c1 -he t test for correlation coefficients * 0d1 -he t test for a re%ression coefficient. &n each of these, the null hypothesis says there is no relationship and the alternative hypothesis says that there is a relationship. (A) &n this case the null hypothesis says that the t o population means (i.e., mu one and mu t o) are e#ual; the alternative hypothesis says that they are not e#ual. (7) &n this case the null hypothesis says that all of the population means are e#ual; the alternative hypothesis says that at least t o of the means are not e#ual. (G) &n this case the null hypothesis says that the population correlation (i.e., rho) is !ero; the alternative hypothesis says that it is not e#ual to !ero. (H) &n this case the null hypothesis says that the population regression coefficient (beta) is !ero, and the alternative says that it is not e#ual to !ero. Iou can e%amples of these null and alternative hypotheses ritten out in symbolic form for cases A, 7, G, and H in the follo ing Table.

SA56: I7S6R- -A896 18.! '6R6