© 2007 Prentice Hall16-1 Some Preliminaries

© 2007 Prentice Hall16-2 Basics of Analysis The process of data analysis Example 1: Gift Catalog Marketer Mails 4 times a year to its customers Company has I million customers on its file Observation DataInformation Encode Analysis

© 2007 Prentice Hall16-3 Example 1 Cataloger would like to know if new customers buy more than old customers? Classify New Customers as anyone who brought within the last twelve months. Analyst takes a sample of 100,000 customers and notices the following.

© 2007 Prentice Hall16-4 Example orders received in the last month 3000 (60%) were from new customers 2000 (40%) were from old customers So it looks like the new customers are doing better

© 2007 Prentice Hall16-5 Example 1 Is there any Catch here!!!!! Data at this gross level, has no discrimination between customers within either group. A customer who bought within the last 11 days is treated exactly similar to a customer who bought within the last 11 months.

© 2007 Prentice Hall16-6 Example 1 Can we use some other variable to distinguish between old and new Customers? Answer: Actual Dollars spent ! What can we do with this variable? Find its Mean and Variation. We might find that the average purchase amount for old customers is two or three times larger than the average among new customers

© 2007 Prentice Hall16-7 Numerical Summaries of data The two basic concepts are the center and the Spread of the data Center of data - Mean, which is given by - Median - Mode

© 2007 Prentice Hall16-8 Numerical Summaries of data Forms of Variation Sum of differences about the mean: Variance: Standard Deviation: Square Root of Variance

© 2007 Prentice Hall16-9 Confidence Intervals In catalog eg, analyst wants to know average purchase amount of customers He draws two samples of 75 customers each and finds the means to be $68 and $122 Since difference is large, he draws another 38 samples of 75 each The mean of means of the 40 samples turns out to be $ How confident should he be of this mean of means?

© 2007 Prentice Hall16-10 Confidence Intervals Analyst calculates the standard deviation of sample means, called Standard Error (SE). It is Basic Premise for confidence Intervals 95 percent of the time the true mean purchase amount lies between plus or minus 1.96 standard errors from the mean of the sample means. C.I. = Mean (+or-) (1.96) * Standard Error

© 2007 Prentice Hall16-11 Confidence Intervals However, if CI is calculated with only one sample then Standard Error of sample mean = Standard deviation of sample Basic Premise for confidence Intervals with one sample 95 percent of the time the true mean lies between plus or minus 1.96 standard errors from the sample means.

© 2007 Prentice Hall16-12 Example 2: Confidence Intervals for response rates You are the marketing analyst for Online Apparel Company You want to run a promotion for all customers on your database In the past you have run many such promotions Historically you needed a 4.5% response for the promotions to break-even You want to test the viability of the current full- scale promotion by running a small test promotion

© 2007 Prentice Hall16-13 Example 2: Confidence Intervals for response rates Test 1,000 names selected at random from a new list. To break-even the list must be expected to have a response rate of 4.5 percent Confidence Interval= Expected Response (+/-) 1.96*SE = p(+/-) 1.96*SE In our case C.I. = 3.22 % to 5.78%. Thus any response between 3.22 and 5.78 % supports hypothesis that true response rate is 4.5%

© 2007 Prentice Hall16-14 Example 2: Confidence Intervals for response rates The list is mailed and actually pulls in 3.5% Thus, the true response rate maybe 4.5% What if the actual rate pulled in were 5% ? Regression towards mean: Phenomenon of test result being different from true result Give more thought to lists whose cutoff rates lie within confidence interval

© 2007 Prentice Hall16-15 Frequency Distribution and Cross-Tabulation © 2007 Prentice Hall15

© 2007 Prentice Hall16-16 Chapter Outline 1) Frequency Distribution 2) Statistics Associated with Frequency Distribution i.Measures of Location ii.Measures of Variability iii.Measures of Shape 3) Cross-Tabulations i.Two Variable Case ii.Three Variable Case iii.General Comments on Cross-Tabulations 4) Statistics for Cross-Tabulation : Chi-Square

© 2007 Prentice Hall16-17 Internet Usage Data Respondent Sex Familiarity Internet Attitude Toward Usage of Internet Number UsageInternetTechnology Shopping Banking Table 15.1

© 2007 Prentice Hall16-18 Frequency Distribution In a frequency distribution, one variable is considered at a time. A frequency distribution for a variable produces a table of frequency counts, percentages, and cumulative percentages for all the values associated with that variable.

© 2007 Prentice Hall16-19 Frequency Distribution of Familiarity with the Internet Table 15.2

© 2007 Prentice Hall16-20 Frequency Histogram Fig Frequency Familiarity 8

© 2007 Prentice Hall16-21 The mean, or average value, is the most commonly used measure of central tendency. The mean,,is given by Where, X i = Observed values of the variable X n = Number of observations (sample size) The mode is the value that occurs most frequently. The mode is a good measure of location when the variable is inherently categorical or has otherwise been grouped into categories. Statistics for Frequency Distribution: Measures of Location X = X i / n  i=1 n X

© 2007 Prentice Hall16-22 The median of a sample is the middle value when the data are arranged in ascending or descending order. If the number of data points is even, the median is the midpoint between the two middle values. The median is the 50th percentile. Statistics for Frequency Distribution: Measures of Location

© 2007 Prentice Hall16-23 The range measures the spread of the data. The variance is the mean squared deviation from the mean. The variance can never be negative. The standard deviation is the square root of the variance. The coefficient of variation is the ratio of the standard deviation to the mean expressed as a percentage, and is a unitless measure of relative variability. CV = s x /X Statistics for Frequency Distribution: Measures of Variability

© 2007 Prentice Hall16-24 Skewness. The tendency of the deviations from the mean to be larger in one direction than in the other. Tendency for one tail of the distribution to be heavier than the other. Kurtosis is a measure of the relative peakedness or flatness of the frequency distribution curve. The kurtosis of a normal distribution is zero. -kurtosis>0, then dist is more peaked than normal dist. -kurtosis<0, then dist is flatter than a normal distribution. Statistics for Frequency Distribution: Measures of Shape

© 2007 Prentice Hall16-25 Skewness of a Distribution Fig Skewed Distribution Symmetric Distribution Mean Median Mode (a) Mean Median Mode (b)

© 2007 Prentice Hall16-26 Cross-Tabulation While a frequency distribution describes one variable at a time, a cross-tabulation describes two or more variables simultaneously. Cross-tabulation results in tables that reflect the joint distribution of two or more variables with a limited number of categories or distinct values, e.g., Table 15.3.

© 2007 Prentice Hall16-27 Gender and Internet Usage Table 15.3 Gender Row Internet Usage MaleFemaleTotal Light (1) Heavy (2) Column Total 15 15

© 2007 Prentice Hall16-28 Two Variables Cross-Tabulation Since two variables have been cross-classified, percentages could be computed either columnwise, based on column totals (Table 15.4), or rowwise, based on row totals (Table 15.5). The general rule is to compute the percentages in the direction of the independent variable, across the dependent variable. The correct way of calculating percentages is as shown in Table 15.4.

© 2007 Prentice Hall16-29 Internet Usage by Gender Table 15.4

© 2007 Prentice Hall16-30 Gender by Internet Usage Table 15.5

© 2007 Prentice Hall16-31 Introduction of a Third Variable in Cross-Tabulation Refined Association between the Two Variables No Association between the Two Variables No Change in the Initial Pattern Some Association between the Two Variables Fig Some Association between the Two Variables No Association between the Two Variables Introduce a Third Variable Original Two Variables

© 2007 Prentice Hall16-32 As can be seen from Table 15.6, 52% (31%) of unmarried (married) respondents fell in the high-purchase category Do unmarried respondents purchase more fashion clothing? A third variable, the buyer's sex, was introduced As shown in Table 15.7, - 60% (25%) of unmarried (married) females fell in the high-purchase category - 40% (35%) of unmarried (married) males fell in the high- purchase category. Unmarried respondents are more likely to fall in the high purchase category than married ones, and this effect is much more pronounced for females than for males. 3 Variables Cross-Tab: Refine an Initial Relationship

© 2007 Prentice Hall16-33 Purchase of Fashion Clothing by Marital Status Table 15.6 Purchase of Fashion Current Marital Status Clothing Married Unmarried High31%52% Low69%48% Column100% Number of respondents

© 2007 Prentice Hall16-34 Purchase of Fashion Clothing by Marital Status and Gender Table 15.7 Purchase of Fashion Clothing Sex Male Female Married Not Married Not Married High35%40%25%60% Low65%60%75%40% Column totals 100% Number of cases

© 2007 Prentice Hall16-35 Table 15.8 shows that 32% (21%) of those with (without) college degrees own an expensive automobile Income may also be a factor In Table 15.9, when the data for the high income and low income groups are examined separately, the association between education and ownership of expensive automobiles disappears, Initial relationship observed between these two variables was spurious. 3 Variables Cross-Tab: Initial Relationship was Spurious

© 2007 Prentice Hall16-36 Ownership of Expensive Automobiles by Education Level Table 15.8 Own Expensive Automobile Education College DegreeNo College Degree Yes No Column totals Number of cases 32% 68% 100% % 79% 100% 750

© 2007 Prentice Hall16-37 Ownership of Expensive Automobiles by Education Level and Income Levels Table 15.9

© 2007 Prentice Hall16-38 Table shows no association between desire to travel abroad and age. In Table 15.11, sex was introduced as the third variable. Controlling for effect of sex, the suppressed association between desire to travel abroad and age is revealed for the separate categories of males and females. Since the association between desire to travel abroad and age runs in the opposite direction for males and females, the relationship between these two variables is masked when the data are aggregated across sex as in Table Variables Cross-Tab: Reveal Suppressed Association

© 2007 Prentice Hall16-39 Desire to Travel Abroad by Age Table 15.10

© 2007 Prentice Hall16-40 Desire to Travel Abroad by Age and Gender Table 15.11

© 2007 Prentice Hall16-41 Consider the cross-tabulation of family size and the tendency to eat out frequently in fast-food restaurants as shown in Table No association is observed. When income was introduced as a third variable in the analysis, Table was obtained. Again, no association was observed. Three Variables Cross-Tabulations No Change in Initial Relationship

© 2007 Prentice Hall16-42 Eating Frequently in Fast-Food Restaurants by Family Size Table 15.12

© 2007 Prentice Hall16-43 Eating Frequently in Fast Food-Restaurants by Family Size and Income Table 15.13

© 2007 Prentice Hall16-44 H 0 : there is no association between the two variables Use chi-square statistic H 0 will be rejected when the calculated value of the test statistic is greater than the critical value of the chi-square distribution Statistics Associated with Cross-Tab: Chi-Square

© 2007 Prentice Hall16-45 Statistics Associated with Cross-Tab: Chi-Square compares the of the observed cell frequencies (f o ) to the frequencies to be expected when there is no association between variables (f e ) The expected frequency for each cell can be calculated by using a simple formula: n r =total number in the row n c =total number in the column n=total sample size

© 2007 Prentice Hall16-46 From Table 3 in the Statistical Appendix, the probability of exceeding a chi-square value of is The calculated chi-square is Since this is less than the critical value of 3.841, the null hypothesis can not be rejected Thus, the association is not statistically significant at the 0.05 level. Statistics for Cross-Tab: Chi-Square