Exploring and Examining Data with Visualizations
•Transferir como PPT, PDF•
1 gostou•477 visualizações
This chapter discusses exploratory data analysis techniques for gaining insights from data. It emphasizes visual representations of data through charts, graphs, and cross-tabulation. Cross-tabulation allows examination of relationships between categorical variables and serves as a framework for statistical testing and decision-making. The chapter also discusses how data exploration identifies unexpected patterns that can lead to innovation. It provides examples of different types of visualizations including bar charts, pie charts, histograms, stem-and-leaf displays, boxplots, and maps. Guidelines are given for interpreting percentages in cross-tabulation tables. The key is to make sense of various data displays to deliver insights.
Recomendados
Recomendados
Mais conteúdo relacionado
Semelhante a Exploring and Examining Data with Visualizations
Semelhante a Exploring and Examining Data with Visualizations (20)
Mais de Dhamo daran
Mais de Dhamo daran (20)
Último
Último (16)
Exploring and Examining Data with Visualizations
- 1. Chapter 16Chapter 16 Exploring, Displaying,Exploring, Displaying, and Examining Dataand Examining Data McGraw-Hill/Irwin Copyright © 2011 by The McGraw-Hill Companies, Inc. All Rights Reserved.
- 2. 16-2 Learning ObjectivesLearning Objectives Understand . . . • That exploratory data analysis techniques provide insights and data diagnostics by emphasizing visual representations of the data. • How cross-tabulation is used to examine relationships involving categorical variables, serves as a framework for later statistical testing, and makes an efficient tool for data visualization and later decision-making.
- 3. 16-3 Research asResearch as Competitive AdvantageCompetitive Advantage “As data availability continues to increase, the importance of identifying/filtering and analyzing relevant data can be a powerful way to gain an information advantage over our competition.” Tom H.C. Anderson founder & managing partner Anderson Analytics, LLC
- 4. 16-4 PulsePoint:PulsePoint: Research RevelationResearch Revelation 65 The percent boost in company revenue created by best practices in data quality.
- 5. 16-5 Researcher Skill Improves DataResearcher Skill Improves Data DiscoveryDiscovery DDW is a global player in research services. As this ad proclaims, you can “push data into a template and get the job done,” but you are unlikely to make discoveries using a template process.
- 6. 16-6 Exploratory Data AnalysisExploratory Data Analysis ConfirmatoryExploratory
- 7. 16-7 Data Exploration, Examination,Data Exploration, Examination, and Analysis in the Researchand Analysis in the Research ProcessProcess
- 8. 16-8 Research Values theResearch Values the UnexpectedUnexpected “It is precisely because the unexpected jolts us out of our preconceived notions, our assumptions, our certainties, that it is such a fertile source of innovation.” Peter Drucker, author Innovation and Entrepreneurship
- 9. 16-9 Frequency of Ad RecallFrequency of Ad Recall Value Label Value Frequency Percent Valid Cumulative Percent Percent
- 17. 16-17 Diagnostics with BoxplotsDiagnostics with Boxplots
- 20. 16-20 Geograph:Geograph: Digital Camera OwnershipDigital Camera Ownership
- 23. 16-23 Guidelines for UsingGuidelines for Using PercentagesPercentages Averaging percentagesAveraging percentages Use of too large percentagesUse of too large percentages Using too small a baseUsing too small a base Percentage decreases can never exceed 100% Percentage decreases can never exceed 100%
- 24. 16-24 Cross-Tabulation with ControlCross-Tabulation with Control and Nested Variablesand Nested Variables
- 25. 16-25 Automatic Interaction DetectionAutomatic Interaction Detection (AID)(AID)
- 26. 16-26 Exploratory Data AnalysisExploratory Data Analysis This Booth Research Services ad suggests that the researcher’s role is to make sense of data displays. Great data exploration and analysis delivers insight from data.
- 27. 16-27 Key TermsKey Terms • Automatic interaction detection (AID) • Boxplot • Cell • Confirmatory data analysis • Contingency table • Control variable • Cross-tabulation • Exploratory data analysis (EDA) • Five-number summary • Frequency table • Histogram • Interquartile range (IQR) • Marginals • Nonresistant statistics • Outliers • Pareto diagram • Resistant statistics • Stem-and-leaf display
- 28. Working with Data Tables McGraw-Hill/Irwin Copyright © 2011 by The McGraw-Hill Companies, Inc. All Rights Reserved.
- 29. 16-29 Original Data TableOriginal Data Table Our grateful appreciation to eMarketer for the use of their table.
- 30. 16-30 Arranged by SpendingArranged by Spending
- 31. 16-31 Arranged byArranged by No. of PurchasesNo. of Purchases
- 32. 16-32 Arranged by Avg. Transaction,Arranged by Avg. Transaction, HighestHighest
- 33. 16-33 Arranged by Avg. Transaction,Arranged by Avg. Transaction, LowestLowest
Notas do Editor
- This chapter presents the use of charts to present data and the initial exploration of data using tools like cross-tabulation.
- See the text Instructors Manual (downloadable from the text website) for ideas for using this research-generated statistic.
- In exploratory data analysis, the researcher has the flexibility to respond to the patterns revealed in the preliminary analysis of the data. Patterns in the collected data guide the data analysis or suggest revisions to the preliminary data analysis plan. This flexibility is an important attribute of this approach. When the researcher is attempting to show causation, confirmatory data analysis is required. Confirmatory data analysis is an analytical process guided by classical statistical inference in its use of significance and confidence.
- Exhibit 16-1 Exhibit 16-1 reminds one of the importance of data visualization as an integral element in the data analysis process and as a necessary step prior to hypothesis testing.
- Exhibit 16-2 A frequency table is a simple device for arraying data. It arrays category codes from lowest value to highest value, with columns for count (frequency), percent, valid percent (percent when missing data is extracted), and cumulative percent. Ad recall, a nominal variable, describes the ads research participants remembered seeing or hearing without being prompted by the researcher or the measurement instrument. Although there are 100 observations, the small number of media placements makes the variable easily tabled. The same data are presented using a pie chart on the next slides.
- Exhibit 16-3 In this slide, the same data are presented in the form of a bar chart.
- Exhibit 16-3, part This portion of Exhibit 16-3 illustrates the observations of ad recall in the form of a pie chart. Data may be more readily understood when presented graphically.
- Exhibit 16-4 Exhibit 16-4 When the variable of interest is measured on an interval-ratio scale and is one with many potential values, these techniques are not particularly informative. Exhibit 16-4, shown in the slide, is a condensed frequency table of the average annual purchases of PrimeSell’s top 50 customers. Only two values, 59.9 and 66, have a frequency greater than 1. Thus, the primary contribution of this table is an ordered list of values. If the table were converted to a bar chart, it would have 48 bars of equal length and two bars with two occurrences.
- Exhibit 16-5 The histogram is the conventional solution for the display of interval-ratio data. Histograms are used when it is possible to group the variable’s values into intervals. A histogram is a graphical bar chart that groups continuous data values into equal intervals, with one bar for each interval. Data analysts find histograms useful for 1) displaying all intervals in a distribution, even those without observed values, and 2) examining the shape of the distribution for skewness, kurtosis, and the modal pattern. The values for the average annual purchases variable presented in Exhibit 16-4 were measured on a ratio scale and are easily grouped. Histograms are not useful for nominal variables like ad recall that has no order to its categories.
- Exhibit 16-6 The stem-and-leaf display is a technique that is closely related to the histogram. It shares some of the histogram’s features but offers several unique advantages. In contrast to histograms, which lose information by grouping data values into intervals, the stem-and-leaf presents actual data values that can be inspected directly, without the use of enclosed bar or asterisks as the representation medium. Visualization is the second advantage of stem-and-leaf displays. The range of values is apparent at a glance, and both shape and spread impressions are immediate. Patterns in the data are easily observed. Each line or row in the display is referred to as a stem, and each piece of information on the stem is called a leaf. In the first stem, there are 12 items (leaves) in the data set whose first digit is 5. 455666788889 representing 54,55,55,56,56,56,57,58,58,58,58,59 The second line shows that there are eight average annual purchase values whose first digit is six. 12366799 representing 61,62,63,66,66,67,69,69
- Exhibit 16-7 Pareto diagrams represent frequency data as a bar chart, ordered from most to least, overlayed with a line graph denoting the cumulative percentage at each variable level. The percentages sum to 100 percent. The data are derived from a multiple-choice-single-response scale, a multiple-choice-multiple-response scale, or frequency counts of words or themes from content analysis. Exhibit 16-7, shown in the slide, depicts an analysis of MindWriter customer complaints as a Pareto diagram
- Exhibit 16-8 The boxplot, or box-and-whisker plot, is another technique used frequently in exploratory data analysis. A boxplot reduces the detail of the stem-and-leaf display and provides a different visual image of the distribution’s location, spread, shape, tail length, and outliers. Boxplots are extensions of the five-number summary of a distribution. This summary consists of the median, the upper and lower quartiles, and the largest and smallest observations. The median and quartiles are used because they are particularly resistant statistics. Resistance is a characteristic that provides insensitivity to localized misbehavior in data. The mean and standard deviation are considered nonresistant statistics, because they are susceptible to the effects of extreme values in the tails of the distribution and do not represent typical values well under conditions of asymmetry. Boxplots may be constructed easily by hand or by computer programs. The ingredients of the plot are The rectangular plot that encompasses 50% of the data values, A center line--marking the median and going through the width of the box, The edges of the box, called hinges, and The whiskers that extend from the right and left hinges to the largest and smallest values. These values may be found within 1.5 times the interquartile range (IQR) from either edge of the box.
- Exhibit 16-9 Exhibit 16-9 summarizes several comparisons that are of help to the analyst. Boxplots are an excellent diagnostic tool, especially when graphed on the same scale. The upper two plots in the exhibit are both symmetric, but one is larger than the other. Larger box widths are sometimes used when the second variable, from the same measurement scale, comes from a larger sample size. The box widths should be proportional to the square root of the sample size, but not all plotting programs account for this. Right- and left-skewed distributions and those with reduced spread are also presented clearly in the plot comparison. Groups may be compared by means of multiple plots.
- In Exhibit 16-10, multiple boxplots compare five sectors of PrimeSell’s customers by their average annual purchases data. The overall impression is one of potential problems for the analyst: unequal variances, skewness, and extreme outliers. Note the similarities of the profiles of finance and retailing in contrast to the high-tech and insurance sectors.
- With mapping, colors and patterns denoting knowledge, attitude, behavior, or demographic data arrays are superimposed over street maps, block-group maps, or county, state, or country maps to help identify the best locations for stores based on demographic, psychographic, and life-stage segmentation data. The PCensus ad points out that determining whether a site has the potential to attract sufficient members of a market and offers facilitating infrastructure and appropriate traffic patterns can be facilitated by mapping.
- This map, developed by American Demographics and Claritas illustrates the penetration of digital cameras by geographic location.
- Exhibit 16-11 Exhibit 16-11 Cross-tabulation is a technique for comparing data from two or more categorical variables. It is used with demographic variables and the study’s target variables. The technique uses tables having rows and columns that correspond to the levels or code values of each variable’s categories. Exhibit 16-11 is an example of a computer-generated cross-tabulation. This table has two rows for gender and two columns for assignment selection. The combination produces four cells. Depending on what you request for each cell, it can contain a count of the cases of the joint classification and also the row, column, and/or the total percentages. The number of row cells and column cells is often used to designate the size of the table, as in this 2 x 2 table. Row and column totals, called marginals, appear at the bottom and right “margins” of the table. When tables are constructed for statistical testing, we call them contingency tables and the test determines if the classification variables are independent of each other. This is discussed in Chapter 20.
- Exhibit 16-12 Percentages serve two purposes in data presentation. They simplify the data by reducing all numbers to a range from 0 to 100. They also translate the data into standard form with a base of 100 for relative comparisons. One can see in Exhibit 16-12 that the percentage of females selected for overseas assignments rose from 15.8 to 22.5 percent of their respective samples. Among all overseas selectees, in the first study, 21.4% were women, while in the second study, 37.5% were women. The tables verify an increase in women with overseas assignments, but we cannot conclude that their gender had anything to do with the increase.
- Percentages are used by virtually everyone dealing with numbers, but these guidelines will help to prevent errors in reporting. Percentiles cannot be averaged unless each is weighted by the size of the group from which it is derived. In other words, a simple average is inappropriate but a weighted average may be used. A large percentage is difficult to understand. For instance, if a 1,000 percent increase is experienced, it is better to describe the increase as a 10-fold increase. Percentages hide the base from which they have been computed. A figure of 60% when contrasted with 30% seems sizable, but there may be only 3 cases in one category and 6 in another. The final guideline shouldn’t happen but does. The higher figure should always be used as the denominator or base. For instance, if a price is reduced from $1 to $.25, the decrease is 75% (75/100).
- Exhibit 16-13 A control variable is a variable introduced to help interpret the relationship between variables. Statistical packages like SPSS have the option of constructing n-way tables with the provision of multiple control variables. Exhibit 16-13 presents an example in which all three variables are handled under the same banner.
- Exhibit 16-14 An advanced variation on n-way tables is automatic interaction detection (AID). AID is a computerized statistical process that requires that the researcher identify a dependent variable and a set of predictors or independent variables. The computer then searches among up to 300 variables for the best single division of the data according to each predictor variable, chooses one, and splits the sample using a statistical test to verify the appropriateness of this choice. Exhibit 16-14 shows the tree diagram that resulted from an AID study of customer satisfaction with MindWriter’s CompleteCare repair service. The initial dependent variable is the overall impression of the repair service. The variable was measured on an interval scale of 1 to 5. The variables that contribute to perceptions of repair effectiveness were also measured on the same scale but were rescaled to ordinal data for this example. The top box shows that 62% of the respondents rated the repair service as excellent. The best predictor of repair effectiveness s “resolution of the problem.”