Download as PPTX, PDF, TXT or read online from Scribd
Download as pptx, pdf, or txt
You are on page 1/ 33
Chapter V
DESCRIPTIVE STATISTICS & INFERENTIAL Descriptive vs Inferential statistics • Descriptive statistics describe a collection of data. E.g., Mean, median, and mode. • Inferential statistics draw inferences about characteristics of a population following examination of a sample. E.g., standard deviation, t-test, anova. Inferential statistics – PARAMETRIC • Parametric statistics : where the population is assumed to fit a parameterized distribution ( most typically the normal distribution ). • Parametric is used when the test variable is measured on an interval or ratio scale Inferential statistics – NON-PARAMETRIC • Sometimes a fit to a distribution doesn’t work. In that case a non-parametric alternative is more likely to detect a difference or lack of difference. • Non-Parametric statistics are statistics where it is not assumed that the population fits any parameterized distributions. • Non-Parametric statistics are typically applied to populations that take on a ranked order. What is Parametric test ? • Parametric tests apply to QuanTitative data ( discrete or continuous ). • Parametric test includes : mean, standard deviation, one sample t-test, two sample t-test, Z-test, ANOVA. • In Parametric tests, the measure of central tendency is : Mean. • To test whether the data are normally distributed, we can use two tests : Kolmogorov– Smirnov test and the Shapiro–Wilk test. Hypothesis testing : PARAMETRIC testing • An Inferential statistical analysis compares sample populations to see whether they are likely to have been drawn from the same population. • For parametric testing of two things : student’s t-test. • For parametric testing of multiple things : Analysis of Variance (ANOVA) What is t-test ? • Use the t-test when you have a small sample size, or if you don’t know the population standard deviation. t-test very similar to a Z-score. • t-test can be used to compare the mean of a sample with a given value. We call this type a “one-sample t-test.” • t-tests can also compare the means of two samples. We call “two-samples t-tests” : including “two independent samples” or “two paired samples.” Types of t-test ? • Independent samples t-tests consider two distinct groups, such as males vs females. • Paired samples t-tests relate to the same set of respondents and thus, occur when respondents are surveyed multiple times. This situation occurs in pre-test/post-test studies designed to measure a variable before and after a treatment. What is Non-parametric test ( also called Non-metric tests ) ? • Non-parametric tests ( also called distribution-free tests ) don’t assume that your data follow a normal distribution. • Non-parametric tests apply to QuaLitative data : nominal or ordinal data ( also called Non-metric independent variables ). • In Non-parametric tests, the Measure of central tendency is : Median. Two cases of Standard Deviation
10 Distribution of Sample Means Examples of Confidence level Example of Significance level • Contractor says that it will take 9 months to construct a house. The house is finished in 9 months and 1 week. The completion time is not 9 months; however it is Not significantly different from the estimated time.
• Local authorities estimate that there are 20,000
people at a concert. Ticket receipts indicate there are 42,000 attendees. This number of 42,000 is significantly different from 20,000. Significance level - Confidence level - p-value • Confidence level is associated with Null hypothesis H0 ( where we accept Null hypothesis H0 ). • Significance level is associated with the areas outside H0 (where we reject Null hypothesis H0 ). • Significance value ( as alpha α ) gives us the Critical Value for testing. • p-value is a quantitative measure of significance. Illustration of confidence & significance Two-tailed hypothesis test One-tailed hypothesis test ( right hand) One-tailed hypothesis test ( left hand) SELECTING METHODS OF DATA ANALYSIS Analyzing QuaLitative data ( Non-metric ) • Nominal : Calculating Frequency Chi-square (kiểm định tần số)
• Ordinal : Calculating Frequency & Median Testing Kolmogorov-Simirnov, Wilcoxon Analyzing QuanTitative data ( metric )
Applying for Interval & Ratio Scale
Calculating Mean Z test, t -test Depending on Parametric vs. Non-parametric • Parametric test Z, t
• Non-parametric Chi-square, Wilcoxon Depending on number of samples • Independent samples e.g., Testing Mean of a population use t for Independent sample
• Dependent samples e.g., Testing difference between 2 Means use t for Paired samples Depending on Number of Variables
• Univariate data analysis
• Bivariate • Multivariate Depending on the correlations btw Variables • Dependence method – Independent variables & Dependent variables; – e.g., Linear regression, Multiple regression, Discriminant analysis, Multivariate analysis of variance. • Interdependence method – When there are no independent nor dependent variables but they depend on each other; – e.g., Exploratory factor analysis, Cluster analysis, Multidimensional scaling. DATA ENTRY Data Coding • Precoding : coding Before interview [ with Closed questions ] • Postcoding : coding After interview [ with Open questions ] • Codes have two digits : – 1st digit : for Variable – 2nd digit : for Answer • Data and codes are described on Code book Data Matrix
• Data Coding Data Matrix Software
• Columns of matrix express Coding of Variables • Rows of matrix express elements of samples • Intersections of columns and rows express the answers Clean missing cells • Mistake from data entry • Mistake from missing field data • To determine the reason : sum up the total number of answers to each variable ( each column ) • If from data entry supplement from questionnaire • If from missing interview – Amend interview – Bypass that element ( reduce sample size ) – Replace by average of some answers or all answers ( SPSS can do fast ) Clean unreasonable data • Mainly due to data entry mistake • Identify by calculating the frequency of occurrence at column ( variable ) • Resolve by checking the questionnaire SUMMARIZING DATA • Statistical Summarization : mean, median, mode, variance, standard deviation, range. • Summarizing by Table: Simple & Cross table • Summarizing by Chart : – Bar chart : – Pie – Line graph – Scattered
