We will be generally concerned with testing hypotheses.  There are many types of hypotheses that may be of interest.  Those listed below will be introduced this semester.  Note that although our tests are usually applied to samples, we are testing hypotheses about population parameters. 

 

Z test

H0: m = m0 where variance (s2) is known

F test

H0: s12 = s22

Chi square test of variance

H0: s2 = s02

Chi square test of independence

H0: the two sets of categories are independent

Chi square goodness of fit test

H0: the pattern of counts follow a given pattern

One-sample t-test

H0: m = m0

Paired t-test

H0: m = m0

Two-sample t-test

H0: m1 = m2

Analysis of Variance (one-way)

H0: m1 = m2 = m3 = m4 = …

Analysis of Variance (two-way)

H01: m11 = m12 = m13 = … & H02: m21 = m22 = m23 = …

Regression

H0: bi = 0

 

1) Test a mean against an hypothesized value. 

a) If the variance is known, use a Z test. 

b) If the variance is estimated from a sample, use a t-test. 

c) The paired t-test is a special case where the two means are for paired observations.  In this case you can take the difference between the two paired observations and test the mean of the pairs as a single sample of differences. 

2) Test a mean against another mean. 

a) A two-sample t test is used to test one mean against another mean. 

b) An Analysis of Variance is used to test between means when there are three or more means (though it can also be to test two means in which case it gives the same result as the t-test). 

c) It can also be used when means from two or more groups are to be compared.  For example, if you wish to compare test results for class (freshman, sophomores, juniors and seniors) and for gender (male and female).  This is a “two-way” or factorial ANOVA. 

3) Testing variances. 

a) A variance can be tested against an hypothesized value with a Chi-square test. 

b) Two variances can be tested for equality with the F test. 

4) Chi square tests can also be used to test certain patterns of count data. 

a) They can be used to determine if numbers of something occur equally frequently in several categories or not.  Patterns other than “equally distributed” can also be tested. 

b) This test can also be used to determine if counts in a two-way table are dependent on the table categories or not. 

5) Regression is a different kind of analysis used to relate (actually correlate) two variables.  It is used to fit the linear equation Yi = b0 + b1Xi

 

Examples and solutions: 

1) According to published reports the production of soybeans in northern Louisiana should average 38 bushels per acre.  A researcher has produced an average of 36 bushels lbs per acre on 24 experimental plots (variance = 9 lb per acre squared).  He wants to know if his production differs significantly from the published mean. 

2) Packing crates used for tomatoes are designed to accommodate a mean size of 3 inches with a standard deviation of no more than 0.3 inches (variance = 0.09 inches squared).  An agronomy student has developed a new variety with a mean of 3 inches and a variance of 0.11 inches squared.  He wants to determine if the variance for his variety exceeds the required standard. 

3) Mendelian genetics dictate that the phenotypic expression of a gene with incomplete dominance should follow a 9:6:1 ratio when heterozygous individuals are crossed.  The results of a genetics experiment on butterfly color patterns show a 127:86:17 ratio.  The investigator wants to determine if this pattern departs significantly from the expected ratio. 

4) A forest products student is developing a new binder for laminated wood products.  He wants to compare the mean strength of laminated wood prepared with his new formulation against the commercial standard.  He prepares 10 pieces of laminated wood with each binder and measures their strength.   The mean and variance of the commercial standard are known values.

5) A sociology student studying college student spending habits.  She has interviewed several hundred students and wants to compare the mean credit card balance for Freshmen, Sophomores, Juniors and Seniors to determine if there are statistically significant differences.