Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
History of statistics wikipedia , lookup
Foundations of statistics wikipedia , lookup
Bootstrapping (statistics) wikipedia , lookup
Taylor's law wikipedia , lookup
Psychometrics wikipedia , lookup
Categorical variable wikipedia , lookup
Omnibus test wikipedia , lookup
Misuse of statistics wikipedia , lookup
GM07: RESEARCH METHODOLOGY UNIT 4-A: DATA ANALYSIS and REPORTING 1. Frequency Distribution 2. Cross Tabulation 3. Hypothesis Testing Organizing Numerical Data • Data in raw form (as collected): 24, 26, 24, 21, 27, 27, 30, 41, 32, 38 • Data in ordered array from smallest to largest: 21, 24, 24, 26, 27, 27, 30, 32, 38, 41 • Stem-and-leaf display: 2 144677 3 028 4 1 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. Organizing Numerical Data Numerical Data Ordered Array 21, 24, 24, 26, 27, 27, 30, 32, 38, 41 41, 24, 32, 26, 27, 27, 30, 24, 38, 21 Frequency Distributions Cumulative Distributions O g ive 120 100 80 60 40 20 0 10 Stem and Leaf Display 2 144677 3 028 4 1 Histograms 20 30 6 5 4 Polygons 3 2 1 0 10 20 30 40 50 60 50 Ogive 7 Tables 40 60 Tabulating Numerical Data: Frequency Distributions • Sort raw data in ascending order: 12, 13, 17, 21, 24, 24, 26, 27, 27, 30, 32, 35, 37, 38, 41, 43, 44, 46, 53, 58 • • • • Find range: 58 - 12 = 46 Select number of classes: 5 (usually between 5 and 15) Compute class interval (width): 10 (46/5 then round up) Determine class boundaries (limits): 10, 20, 30, 40, 50, 60 • Compute class midpoints: 15, 25, 35, 45, 55 • Count observations & assign to classes Frequency Distributions, Relative Frequency Distributions and Percentage Distributions Data in ordered array: 12, 13, 17, 21, 24, 24, 26, 27, 27, 30, 32, 35, 37, 38, 41, 43, 44, 46, 53, 58 Class 10 but under 20 20 but under 30 30 but under 40 40 but under 50 50 but under 60 Total Relative Frequency Frequency Percentage 3 6 5 4 2 20 .15 .30 .25 .20 .10 1 15 30 25 20 10 100 Graphing Numerical Data: The Histogram Data in ordered array: 12, 13, 17, 21, 24, 24, 26, 27, 27, 30, 32, 35, 37, 38, 41, 43, 44, 46, 53, 58 Frequency Histogram 7 6 5 4 3 2 1 0 6 5 3 2 0 5 Class Boundaries No Gaps Between Bars 4 0 15 25 36 45 Class Midpoints 55 More Graphing Numerical Data: The Frequency Polygon Data in ordered array: 12, 13, 17, 21, 24, 24, 26, 27, 27, 30, 32, 35, 37, 38, 41, 43, 44, 46, 53, 58 Frequenc y 7 6 5 4 3 2 1 0 5 15 25 36 45 55 Class Midpoints M ore Tabulating Numerical Data: Cumulative Frequency Data in ordered array: 12, 13, 17, 21, 24, 24, 26, 27, 27, 30, 32, 35, 37, 38, 41, 43, 44, 46, 53, 58 Class 10 but under 20 20 but under 30 30 but under 40 40 but under 50 50 but under 60 Cumulative Frequency 3 9 14 18 20 Cumulative % Frequency 15 45 70 90 100 Graphing Numerical Data: The Ogive (Cumulative % Polygon) Data in ordered array: 12, 13, 17, 21, 24, 24, 26, 27, 27, 30, 32, 35, 37, 38, 41, 43, 44, 46, 53, 58 Ogive 100 80 60 40 20 0 10 20 30 40 50 60 Class Boundaries (Not Midpoints) Graphing Bivariate Numerical Data (Scatter Plot) Mutual Funds Scatter Plot Total Year to Date Return (%) 40 30 20 10 0 0 10 20 30 Net Asset Values 40 Tabulating and Graphing Categorical Data:Univariate Data Categorical Data Tabulating Data The Summary Table Graphing Data Pie Charts Bar Charts Pareto Diagram Summary Table (for an Investor’s Portfolio) Investment Category Amount Percentage (in thousands ) Stocks Bonds CD Savings Total 46.5 32 15.5 16 110 Variables are Categorical 42.27 29.09 14.09 14.55 100 Graphing Categorical Data: Univariate Data Categorical Data Graphing Data Tabulating Data The Summary Table Pie Charts CD Pareto Diagram S a vi n g s Bar Charts B onds S to c k s 0 10 20 30 40 50 45 120 40 100 35 30 80 25 60 20 15 40 10 20 5 0 0 S to c k s B onds S a vi n g s CD Bar Chart (for an Investor’s Portfolio) Investor's Portfolio Savings CD Bonds Stocks 0 10 20 30 Amount in K$ 40 50 Pie Chart (for an Investor’s Portfolio) Amount Invested in K$ Savings 15% Stocks 42% CD 14% Bonds 29% Percentages are rounded to the nearest percent. Pareto Diagram Axis for bar chart shows % invested in each category 45% 100% 40% 90% 80% 35% 70% 30% 60% 25% 50% 20% 40% 15% 30% 10% 20% 5% 10% 0% 0% Stocks Bonds Savings CD Axis for line graph shows cumulative % invested Tabulating and Graphing Bivariate Categorical Data • Contingency tables: investment in thousands Investment Category Investor A Stocks Bonds CD Savings 46.5 32 15.5 16 Total 110 Investor B Investor C Total 55 44 20 28 27.5 19 13.5 7 129 95 49 51 147 67 324 Tabulating and Graphing Bivariate Categorical Data • Side by side charts C o m p arin g In vesto rs S avings CD B onds S toc k s 0 10 Inves tor A 20 30 Inves tor B 40 50 Inves tor C 60 Principles of Graphical Excellence • Presents data in a way that provides substance, statistics and design • Communicates complex ideas with clarity, precision and efficiency • Gives the largest number of ideas in the most efficient manner • Almost always involves several dimensions • Tells the truth about the data “Chart Junk” Bad Presentation Good Presentation Minimum Wage 1960: $1.00 Minimum Wage 4 $ 1970: $1.60 2 1980: $3.10 0 1990: $3.80 1960 1970 1980 1990 No Relative Basis Bad Presentation Good Presentation A’s received by Freq. students. 300 200 30 % 10 0 FR SO JR SR A’s received by students. FR SO JR SR FR = Freshmen, SO = Sophomore, JR = Junior, SR = Senior Compressing Vertical Axis Bad Presentation Good Presentation Quarterly Sales 200 $ Quarterly Sales 50 100 25 0 0 Q1 Q2 Q3 Q4 $ Q1 Q2 Q3 Q4 No Zero Point on Vertical Axis Bad Presentation Good Presentation Monthly Sales 45 $ Monthly Sales 42 39 45 42 39 $ 36 36 J F M A M J 0 Graphing the first six months of sales. J F M A M J Statistics for Frequency Distribution • Measures of central tendency – Mean, median, mode, geometric mean • Quartile • Measure of variation – Range, Interquartile range, variance and standard deviation, coefficient of variation • Measure of Shape – Symmetric, skewed, using box-and-whisker plots Measures of Central Tendency Central Tendency Average Median Mode n X X i 1 n i 1 N Geometric Mean X G X1 X 2 X n 1/ n N X i i Mean (Arithmetic Mean) • Mean (arithmetic mean) of data values – Sample mean Sample Size n X X i 1 i n X1 X 2 n – Population mean N X i 1 N i Xn Population Size X1 X 2 N XN Mean (Arithmetic Mean) (continued) • The most common measure of central tendency • Affected by extreme values (outliers) 0 1 2 3 4 5 6 7 8 9 10 Mean = 5 0 1 2 3 4 5 6 7 8 9 10 12 14 Mean = 6 Median • Robust measure of central tendency • Not affected by extreme values 0 1 2 3 4 5 6 7 8 9 10 Median = 5 0 1 2 3 4 5 6 7 8 9 10 12 14 Median = 5 • In an ordered array, the median is the “middle” number – If n or N is odd, the median is the middle number – If n or N is even, the median is the average of the two middle numbers Mode • • • • • • A measure of central tendency Value that occurs most often Not affected by extreme values Used for either numerical or categorical data There may be no mode There may be several modes 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Mode = 9 0 1 2 3 4 5 6 No Mode Geometric Mean • Useful in the measure of rate of change of a variable over time X G X1 X 2 Xn 1/ n • Geometric mean rate of return – Measures the status of an investment over time RG 1 R1 1 R2 1 Rn 1/ n 1 Quartiles • Split Ordered Data into 4 Quarters 25% 25% Q1 25% Q2 • Position of i-th Quartile 25% Q3 i n 1 Qi 4 Data in Ordered Array: 11 12 13 16 16 17 18 21 22 1 9 1 Position of Q1 2.5 4 Q1 12 13 12.5 2 • Q1and Q3 Are Measures of Noncentral Location • Q2 = Median, A Measure of Central Tendency Measures of Variation Variation Variance Range Population Variance Sample Variance Interquartile Range Standard Deviation Population Standard Deviation Sample Standard Deviation Coefficient of Variation Range • Measure of variation • Difference between the largest and the smallest observations: Range X Largest X Smallest • Ignores the way in which data are distributed Range = 12 - 7 = 5 Range = 12 - 7 = 5 7 8 9 10 11 12 7 8 9 10 11 12 Interquartile Range • Measure of variation • Also known as midspread – Spread in the middle 50% • Difference between the first and third quartiles Data in Ordered Array: 11 12 13 16 16 17 17 18 21 Interquartile Range Q3 Q1 17.5 12.5 5 • Not affected by extreme values Variance • Important measure of variation • Shows variation about the mean – Sample variance: n S 2 – Population variance: X i 1 n 1 N 2 X i X i 1 i N 2 2 Standard Deviation • Most important measure of variation • Shows variation about the mean • Has the same units as the original data – Sample standard deviation: S – Population standard deviation: n X i 1 X i 2 n 1 N X i 1 i N 2 Comparing Standard Deviations Data A 11 12 13 14 15 16 17 18 19 20 21 Mean = 15.5 s = 3.338 Data B 11 12 13 14 15 16 17 18 19 20 21 Mean = 15.5 s = .9258 Data C 11 12 13 14 15 16 17 18 19 20 21 Mean = 15.5 s = 4.57 Coefficient of Variation • Measures relative variation • Always in percentage (%) • Shows variation relative to mean • Is used to compare two or more sets of data measured in different units • S CV X 100% Comparing Coefficient of Variation • Stock A: – Average price last year = $50 – Standard deviation = $5 • Stock B: – Average price last year = $100 – Standard deviation = $5 • Coefficient of variation: – Stock A: S CV X $5 100% 100% 10% $50 – Stock B: S CV X $5 100% 100% 5% $100 Shape of a Distribution • Describes how data is distributed • Measures of shape – Symmetric or skewed Left-Skewed Mean < Median < Mode Symmetric Mean = Median =Mode Right-Skewed Mode < Median < Mean Exploratory Data Analysis • Box-and-whisker plot – Graphical display of data using 5-number summary X smallest Q 1 4 6 Median( Q2) 8 Q3 10 Xlargest 12 Distribution Shape and Box-and-Whisker Plot Left-Skewed Q1 Q2 Q3 Symmetric Q1Q2Q3 Right-Skewed Q1 Q2 Q3 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 Gender and Internet Usage Gender Internet Usage Male Female Row Total Light (1) 5 10 15 Heavy (2) 10 5 15 Column Total 15 15 Two Variables Cross-Tabulation • Since two variables have been cross-classified, percentages could be computed either columnwise, based on column totals, or rowwise, based on row totals . • 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 next slide. Internet Usage by Gender Gender Internet Usage Male Female Light 33.3% 66.7% Heavy 66.7% 33.3% Column total 100% 100% Gender by Internet Usage Internet Usage Gender Light Heavy Total Male 33.3% 66.7% 100.0% Female 66.7% 33.3% 100.0% Introduction of a Third Variable in Cross-Tabulation Original Two Variables Some Association between the Two Variables Introduce a Third Variable Refined Association No Association between the Two between the Two Variables Variables No Association between the Two Variables Introduce a Third Variable No Change in the Initial Pattern Some Association between the Two Variables Three Variables Cross-Tabulation Refine an Initial Relationship The introduction of a third variable can result in four possibilities: • As can be seen from, 52% of unmarried respondents fell in the high-purchase category, as opposed to 31% of the married respondents. Before concluding that unmarried respondents purchase more fashion clothing than those who are married, a third variable, the buyer's sex, was introduced into the analysis. • As shown in the table, in the case of females, 60% of the unmarried fall in the high-purchase category, as compared to 25% of those who are married. On the other hand, the percentages are much closer for males, with 40% of the unmarried and 35% of the married falling in the high purchase category. • Hence, the introduction of sex (third variable) has refined the relationship between marital status and purchase of fashion clothing (original variables). 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. Purchase of Fashion Clothing by Marital Status Purchase of Fashion Clothing Current Marital Status Married Unmarried High 31% 52% Low 69% 48% Column 100% 100% 700 300 Number of respondents Purchase of Fashion Clothing by Marital Status Purchase of Fashion Clothing Sex Male Female Married Not Married Married Not Married High 35% 40% 25% 60% Low 65% 60% 75% 40% Column totals Number of cases 100% 100% 100% 100% 400 120 300 180 Three Variables Cross-Tabulation Initial Relationship was Spurious • Table shows that 32% of those with college degrees own an expensive automobile, as compared to 21% of those without college degrees. Realizing that income may also be a factor, the researcher decided to reexamine the relationship between education and ownership of expensive automobiles in light of income level. • In Table, the percentages of those with and without college degrees who own expensive automobiles are the same for each of the income groups. When the data for the high income and low income groups are examined separately, the association between education and ownership of expensive automobiles disappears, indicating that the initial relationship observed between these two variables was spurious. Ownership of Expensive Automobiles by Education Level Own Expensive Automobile Education College Degree No College Degree Yes 32% 21% No 68% 79% Column totals 100% 100% Number of cases 250 750 Ownership of Expensive Automobiles by Education Level and Income Levels Income Own Expensive Automobile Low Income High Income College Degree No College Degree College Degree No College Degree Yes 20% 20% 40% 40% No 80% 80% 60% 60% 100% 100% 100% 100% 100 700 150 50 Column totals Number of respondents Three Variables Cross-Tabulation Reveal Suppressed Association • Table shows no association between desire to travel abroad and age. • When sex was introduced as the third variable, Table was obtained. Among men, 60% of those under 45 indicated a desire to travel abroad, as compared to 40% of those 45 or older. The pattern was reversed for women, where 35% of those under 45 indicated a desire to travel abroad as opposed to 65% of those 45 or older. • 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. • But when the effect of sex is controlled, as in Table , the suppressed association between desire to travel abroad and age is revealed for the separate categories of males and females. Desire to Travel Abroad by Age Desire to Travel Abroad Age Less than 45 45 or More Yes 50% 50% No 50% 50% Column totals 100% 100% 500 500 Number of respondents Desire to Travel Abroad by Age and Gender Desir e to Tr avel Abr oad Sex Male Age Female Age < 45 >=45 <45 >=45 Yes 60% 40% 35% 65% No 40% 60% 65% 35% 100% 100% 100% 100% 300 300 200 200 Column totals Number of Cases Three Variables Cross-Tabulations No Change in Initial Relationship • 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. Eating Frequently in Fast-Food Restaurants by Family Size Eat Frequently in FastFood Restaurants Family Size Small Large Yes 65% 65% No 35% 35% Column totals 100% 100% 500 500 Number of cases Eating Frequently in Fast Food-Restaurants by Family Size and Income Income Eat Frequently in FastFood Restaurants Low Family size Small Large Yes 65% 65% No 35% 35% Column totals 100% 100% Number of respondents 250 250 High Family size Small Large 65% 65% 35% 35% 100% 100% 250 250 Statistics Associated with Cross-Tabulation Chi-Square • To determine whether a systematic association exists, the probability of obtaining a value of chi-square as large or larger than the one calculated from the cross-tabulation is estimated. • An important characteristic of the chi-square statistic is the number of degrees of freedom (df) associated with it. That is, df = (r - 1) x (c -1). • The null hypothesis (H0) of no association between the two variables will be rejected only when the calculated value of the test statistic is greater than the critical value of the chi-square distribution with the appropriate degrees of freedom, as shown. Statistics Associated with Cross-Tabulation Phi Coefficient • The phi coefficient (f) is used as a measure of the strength of association in the special case of a table with two rows and two columns (a 2 x 2 table). • The phi coefficient is proportional to the square root of the chi-square statistic c2 f= n • It takes the value of 0 when there is no association, which would be indicated by a chi-square value of 0 as well. When the variables are perfectly associated, phi assumes the value of 1 and all the observations fall just on the main or minor diagonal. Statistics Associated with Cross-Tabulation Contingency Coefficient • While the phi coefficient is specific to a 2 x 2 table, the contingency coefficient (C) can be used to assess the strength of association in a table of any size. C= c2 c2 + n • The contingency coefficient varies between 0 and 1. • The maximum value of the contingency coefficient depends on the size of the table (number of rows and number of columns). For this reason, it should be used only to compare tables of the same size. Statistics Associated with Cross-Tabulation Cramer’s V • Cramer's V is a modified version of the phi correlation coefficient, f, and is used in tables larger than 2 x 2. V= or V= f2 min (r-1), (c-1) c2/n min (r-1), (c-1) Statistics Associated with Cross-Tabulation Lambda Coefficient • Asymmetric lambda measures the percentage improvement in predicting the value of the dependent variable, given the value of the independent variable. • Lambda also varies between 0 and 1. A value of 0 means no improvement in prediction. A value of 1 indicates that the prediction can be made without error. This happens when each independent variable category is associated with a single category of the dependent variable. • Asymmetric lambda is computed for each of the variables (treating it as the dependent variable). • A symmetric lambda is also computed, which is a kind of average of the two asymmetric values. The symmetric lambda does not make an assumption about which variable is dependent. It measures the overall improvement when prediction is done in both directions. Other Statistics Associated with Cross-Tabulation • Other statistics like tau b, tau c, and gamma are available to measure association between two ordinal-level variables. Both tau b and tau c adjust for ties. • Tau b is the most appropriate with square tables in which the number of rows and the number of columns are equal. Its value varies between +1 and -1. • For a rectangular table in which the number of rows is different than the number of columns, tau c should be used. • Gamma does not make an adjustment for either ties or table size. Gamma also varies between +1 and -1 and generally has a higher numerical value than tau b or tau c. Cross-Tabulation in Practice While conducting cross-tabulation analysis in practice, it is useful to proceed along the following steps. 1. Test the null hypothesis that there is no association between the variables using the chi-square statistic. If you fail to reject the null hypothesis, then there is no relationship. 2. If H0 is rejected, then determine the strength of the association using an appropriate statistic (phi-coefficient, contingency coefficient, Cramer's V, lambda coefficient, or other statistics), as discussed earlier. 3. If H0 is rejected, interpret the pattern of the relationship by computing the percentages in the direction of the independent variable, across the dependent variable. 4. If the variables are treated as ordinal rather than nominal, use tau b, tau c, or Gamma as the test statistic. If H0 is rejected, then determine the strength of the association using the magnitude, and the direction of the relationship using the sign of the test statistic. Hypothesis Testing In statistics, a hypothesis is a claim or statement about a property of a population. A hypothesis test (or test of significance) is a standard procedure for testing a claim about a property of a population. Rare Event Rule for Inferential Statistics If, under a given assumption, the probability of a particular observed event is exceptionally small, we conclude that the assumption is probably not correct Steps Involved in Hypothesis Testing Formulate H0 and H1 Select Appropriate Test Choose Level of Significance Collect Data and Calculate Test Statistic Determine Probability Associated with Test Statistic Compare with Level of Significance, Determine Critical Value of Test Statistic TSCR Determine if TSCR falls into (Non) Rejection Region Reject or Do not Reject H0 Draw Research Conclusion Note about Identifying H0 and H1 Definitions Critical Region The critical region (or rejection region) is the set of all values of the test statistic that cause us to reject the null hypothesis. Significance Level The significance level (denoted by ) is the probability that the test statistic will fall in the critical region when the null hypothesis is actually true. Common choices for are 0.05, 0.01, and 0.10. Critical Value A critical value is any value that separates the critical region (where we reject the null hypothesis) from the values of the test statistic that do not lead to rejection of the null hypothesis, the sampling distribution that applies, and the significance level . The critical value of z = 1.96 corresponds to a significance level of = 0.05. Type I & Type II Errors A Type I error is the mistake of rejecting the null hypothesis when it is true. The symbol (alpha) is used to represent the probability of a type I error. A Type II error is the mistake of failing to reject the null hypothesis when it is false. The symbol (beta) is used to represent the probability of a type II error. Controlling Type I and Type II Errors For any fixed , an increase in the sample size n will cause a decrease in For any fixed sample size n , a decrease in will cause an increase in . Conversely, an increase in will cause a decrease in . To decrease both and , increase the sample size. Conclusions in Hypothesis Testing We always test the null hypothesis. 1. Reject the H0 2. Fail to reject the H0 Two-tailed, Right-tailed, Left-tailed Tests • The tails in a distribution are the extreme regions bounded by critical values. Two-tailed Test H0: = H1: is divided equally between the two tails of the critical region Means less than or greater than Right-tailed Test H0: = H1: > Points Right Left-tailed Test H0: = H1: < Points Left Decision Criterion • Traditional method: Reject H0 if the test statistic falls within the critical region. Fail to reject H0 if the test statistic does not fall within the critical region. • P-value method: Reject H0 if P-value (where is the significance level, such as 0.05). Fail to reject H0 if P-value > . • Another option: Instead of using a significance level such as 0.05, simply identify the P-value and leave the decision to the reader. Decision Criterion • Confidence Intervals: Because a confidence interval estimate of a population parameter contains the likely values of that parameter, reject a claim that the population parameter has a value that is not included in the confidence interval. P-Value The P-value (or p-value or probability value) is the probability of getting a value of the test statistic that is at least as extreme as the one representing the sample data, assuming that the null hypothesis is true. The null hypothesis is rejected if the P-value is very small, such as 0.05 or less. Example: Finding P-values. Wording of Final Conclusion Accept versus Fail to Reject Some texts use “accept the null hypothesis.” We are not proving the null hypothesis. The sample evidence is not strong enough to warrant rejection (such as not enough evidence to convict a suspect). Comprehensive Hypothesis Test Statistical Tests Parametric Non-Parametric Interval or ratio Scaled Data Assumption about population probability distribution Nominal or ordinal data No assumption about population probability distribution Example Z, t, F test etc. Example χ2, sign, Wilcoxon Signed-Rank, Kruskal-Wallis Test etc. A Classification of Hypothesis Testing Procedures for Examining Differences Hypothesis Tests Non-parametric Tests (Nonmetric Tests) Parametric Tests (Metric Tests) One Sample * t test * Z test Two or More Samples Independent Paired Samples Samples * Two-Group t test * Z test * Paired t test One Sample * Chi-Square * K-S * Runs * Binomial Two or More Samples Independent Samples * Chi-Square * Mann-Whitney * Median * K-S Paired Samples * Sign * Wilcoxon * McNemar * Chi-Square A Broad Classification of Hypothesis Tests Hypothesis Tests Tests of Differences Tests of Association Distributions Means Proportions Median/ Rankings Applications of Z – test (n>30) • • • • Test of significance for single mean Test of significance for difference of means Test of significance for difference of standard deviation (s.d.) Testing a Claim about a Proportion • Testing difference of Two proportions Test of significance for single mean Z X Where n • X Sample Mean • Population mean • Population standard deviation(s.d.) • n Sample size NOTE: If population standard deviation(s.d.) is unknown then estimated sample standard deviation s will be used. 1 S (X X ) n n 95% confidence interval for is i 1 X 1.96 99% confidence interval for is X 2.57 i n n 2 Test of significance for difference of means X X Z 1 2 1 n 1 2 2 2 n 2 Where X 1, X 2are the means of first and second sample 1, 2 are standard deviations of samples n1, n 2 are the sample size of first sample & second sample 2 2not known then 1 • IF 12 and and are 2 Z X X s s n n 1 2 2 1 1 2 2 2 Test of significance for difference of standard deviation (s.d.) Z s s 1 2 2 1 2n 2 2n 1 • IF 1and 2 2 not known then are 2 s s Z s s 2n 2n 1 2 2 2 1 2 1 2 Testing a Claim about a Proportion n = number of trials p = x (sample proportion) n p = population proportion (used in the null hypothesis) q=1–p Z ˆp p pq n Testing difference of Two proportions P1 P 2 Z SE Where n1 n2 SE p [1 p ] n1 * n2 * * Note that p* is the combined two sample proportion weighted by the two sample sizes (n1 * p1 ) (n2 * p2 ) P n1 n2 * Applications of t – test(n≤30) • • • • Test the significance of the mean of a random sample Test the difference between means of two samples (Independent Samples) Test the difference between means of two samples (Dependent Samples) Test of significance of an observed correlation coefficients Test the significance of the mean of a random sample ( X ) n t s Where X Sample Mean Population mean S standard deviation of the sample= n Sample size Degree of freedom(d.f.)= n-1 95% confidence interval for is X t( 0.05,n1) n 1 n 2 ( X X ) i n 1 i 1 Test of significance for difference of means Where X 1 X 2 n1n2 t ( ) S n1 n2 X 1 X 2 1 2 n1 n 2 are the mean, standard deviation and sample size of first and second sample n1 S combined standard deviation S IF s1, s2 and n1 n2 are given d.f. = n1+ n2 -2 S (X i 1 n2 X1) (X i X 2 )2 2 i i 1 n1 n2 2 ( n1 1) s12 ( n 2 1) s 22 n1 n 2 2 Test the difference between means of two samples (Dependent Samples) d t S n Where d is the mean of the differences n is the number of paired observations S is standard deviation of differences d.f. = n – 1 1 n 2 ( d d ) i n 1 i 1 Test of significance of an observed correlation coefficients t r 1 r 2 n2 Where r is correlation coefficient n is sample size d.f. = n – 2 Two Independent Samples F Test An F test of sample variance may be performed if it is not known whether the two populations have equal variance. In this case, the hypotheses are: H0:12 = 22 H1:12 22 Two Independent Samples F Statistic The F statistic is computed from the sample variances as follows s12 F(n1-1),(n2-1) = 2 s 2 where n1 n2 n1-1 n2-1 s12 s22 = size of sample 1 = size of sample 2 = degrees of freedom for sample 1 = degrees of freedom for sample 2 = sample variance for sample 1 = sample variance for sample 2 suppose we wanted to determine whether Internet usage was different for males as compared to females. A two-independentsamples t test was conducted. Nonparametric Tests Nonparametric tests are used when the independent variables are nonmetric. Like parametric tests, nonparametric tests are available for testing variables from one sample, two independent samples, or two related samples. Non-parametric Test • • • • • • • • Chi-Square Test Binomial Runs 1-Samples K-S 2-Independent samples K-Independent Samples 2-Dependent Samples K-Dependent Samples Applications of χ2 Test 1. Goodness of Fit 2. Contingency Analysis (or Test of Independence) 3. Test of population variance Goodness of Fit n Oi Ei i 1 Ei c 2 2 Where Oi and Ei are Observed and expected frequencies Degree of freedom(d.f.) = n-1 Note: • No Ei should be less than 5, if so cell(s) must be combined and d.f. should be reduced accordingly • If some parameters are calculated from Oi to calculate Ei e.g. mean or standard deviation etc then d.f. should be reduced by 1 for each such parameters. Contingency Analysis (or Test of Independence) n Oi Ei i 1 Ei c 2 2 Where Oi and Ei are Observed and expected frequencies Degree of freedom = (rows-1)(column-1) Note: When Degree of freedom is 1 AND N<50, adjust χ2 by Yates's Correction Factor i.e. n c 2 i 1 O E i i 0.5 2 Ei Unless O E 0.5 in such case original (Oi – Ei)2 term is preserved i i Test of population variance (n 1) S c 2 2 Where S2 is sample variance σ2 is population variance Degree of freedom = (n-1) n number of observations 2 Nonparametric Tests One Sample Sometimes the researcher wants to test whether the observations for a particular variable could reasonably have come from a particular distribution, such as the normal, uniform, or Poisson distribution. The Kolmogorov-Smirnov (K-S) one-sample test is one such goodness-of-fit test. The K-S compares the cumulative distribution function for a variable with a specified distribution. Ai denotes the cumulative relative frequency for each category of the theoretical (assumed) distribution, and Oi the comparable value of the sample frequency. The K-S test is based on the maximum value of the absolute difference between Ai and Oi. The test statistic is K = Max A i - Oi Nonparametric Tests One Sample • The decision to reject the null hypothesis is based on the value of K. The larger the K is, the more confidence we have that H0 is false. For = 0.05, thecritical value of K for large samples (over 35) is given by 1.36/ Alternatively, K can n be transformed into a normally distributed z statistic and its associated probability determined. • In the context of the Internet usage example, suppose we wanted to test whether the distribution of Internet usage was normal. A K-S one-sample test is conducted, yielding the data shown in Table indicates that the probability of observing a K value of 0.222, as determined by the normalized z statistic, is 0.103. Since this is more than the significance level of 0.05, the null hypothesis can not be rejected, leading to the same conclusion. Hence, the distribution of Internet usage does not deviate significantly from the normal distribution. K-S One-Sample Test for Normality of Internet Usage Test Distribution - Normal Mean: Standard Deviation: Cases: 6.600 4.296 30 Most Extreme Differences Absolute Positive Negative 0.222 0.222 -0.142 K-S z 1.217 2-Tailed p 0.103 Nonparametric Tests One Sample • The chi-square test can also be performed on a single variable from one sample. In this context, the chi-square serves as a goodness-of-fit test. • The runs test is a test of randomness for the dichotomous variables. This test is conducted by determining whether the order or sequence in which observations are obtained is random. • The binomial test is also a goodness-of-fit test for dichotomous variables. It tests the goodness of fit of the observed number of observations in each category to the number expected under a specified binomial distribution. Nonparametric Tests Two Independent Samples • When the difference in the location of two populations is to be compared based on observations from two independent samples, and the variable is measured on an ordinal scale, the Mann-Whitney U test can be used. • In the Mann-Whitney U test, the two samples are combined and the cases are ranked in order of increasing size. • The test statistic, U, is computed as the number of times a score from sample or group 1 precedes a score from group 2. • If the samples are from the same population, the distribution of scores from the two groups in the rank list should be random. An extreme value of U would indicate a nonrandom pattern, pointing to the inequality of the two groups. • For samples of less than 30, the exact significance level for U is computed. For larger samples, U is transformed into a normally distributed z statistic. This z can be corrected for ties within ranks. Nonparametric Tests Two Independent Samples • We examine again the difference in the Internet usage of males and females. This time, though, the Mann-Whitney U test is used. The results are given in Table . • One could also use the cross-tabulation procedure to conduct a chisquare test. In this case, we will have a 2 x 2 table. One variable will be used to denote the sample, and will assume the value 1 for sample 1 and the value of 2 for sample 2. The other variable will be the binary variable of interest. • The two-sample median test determines whether the two groups are drawn from populations with the same median. It is not as powerful as the Mann-Whitney U test because it merely uses the location of each observation relative to the median, and not the rank, of each observation. • The Kolmogorov-Smirnov two-sample test examines whether the two distributions are the same. It takes into account any differences between the two distributions, including the median, dispersion, and skewness. Mann-Whitney U - Wilcoxon Rank Sum W Test Internet Usage by Gender Sex Mean Rank Cases 20.93 10.07 15 15 Male Female Total 30 U 31.000 W 151.000 z -3.406 Corrected for ties 2-tailed p 0.001 Note U = Mann-Whitney test statistic W= Wilcoxon W Statistic z = U transformed into normally distributed z statistic. Nonparametric Tests Paired Samples • The Wilcoxon matched-pairs signed-ranks test analyzes the differences between the paired observations, taking into account the magnitude of the differences. • It computes the differences between the pairs of variables and ranks the absolute differences. • The next step is to sum the positive and negative ranks. The test statistic, z, is computed from the positive and negative rank sums. • Under the null hypothesis of no difference, z is a standard normal variate with mean 0 and variance 1 for large samples. Nonparametric Tests Paired Samples • The example considered for the paired t test, whether the respondents differed in terms of attitude toward the Internet and attitude toward technology, is considered again. Suppose we assume that both these variables are measured on ordinal rather than interval scales. Accordingly, we use the Wilcoxon test. • The sign test is not as powerful as the Wilcoxon matched-pairs signed-ranks test as it only compares the signs of the differences between pairs of variables without taking into account the ranks. • In the special case of a binary variable where the researcher wishes to test differences in proportions, the McNemar test can be used. Alternatively, the chi-square test can also be used for binary variables. Wilcoxon Matched-Pairs Signed-Rank Test Internet with Technology (Technology - Internet) Cases -Ranks 23 +Ranks 1 Ties 6 Total 30 z = -4.207 Mean rank 12.72 7.50 2-tailed p = 0.0000 A Summary of Hypothesis Tests Related to Differences Sample Application Level of Scaling One Sample Proportion Metric One Sample Distributions Nonmetric Test/Comments Z test K-S and chi-square for goodness of fit Runs test for randomness Binomial test for goodness of fit for dichotomous variables One Sample Means Metric t test, if variance is unknown z test, if variance is known A Summary of Hypothesis Tests Related to Differences Two Independent Samples Two independent samples Distributions Nonmetric K-S two-sample test for examining the equivalence of two distributions Two independent samples Means Metric Two-group t test F test for equality of variances Two independent samples Proportions Metric Nonmetric z test Chi-square test Two independent samples Rankings/Medians Nonmetric Mann-Whitney U test is more powerful than the median test A Summary of Hypothesis Tests Related to Differences Paired Samples Paired samples Means Metric Paired t test Paired samples Proportions Nonmetric Paired samples Rankings/Medians Nonmetric McNemar test for binary variables Chi-square test Wilcoxon matched-pairs ranked-signs test is more powerful than the sign test