Understanding data

Data Analysis September 14, 2010

Do data have structures? Research data have underlying structures Understanding structures makes the data manageable, interesting and useful

Patterns and Exceptions Patterns Data may show, buyers of brand A are less affluent than buyers of brand B Younger consumers buy a brand more often than older consumers

Exceptions Brand A is bought mostly by women except in Hyderabad where it is bought mostly by men

Example – Two variables – Bivariate Analysis Ratings of a brand X on a 10-point scales Men Women Flavor 8.5 8.7 Taste 7.3 7.6 Color 6.4 6.1 Strength 7.7 7.5 Consistency 6.0 6.3 Package 6.1 3.2 Texture 4.6 5.9

Looking for Structures STORES Statement A B C D E Mean Friendly staff 87 84 90 93 91 89 Knowledgeable staff 82 81 88 89 86 85 Helpful staff 78 74 77 83 84 79 Variety of Items 82 78 85 85 83 83 Satisfied Customers 87 80 89 92 94 88 Meets cust Expec 76 71 77 83 81 78 Deli v on promises 80 76 82 90 87 83 Meets needs 81 75 83 88 88 83 Low credit int 72 65 66 75 79 71 Everyday Shopng 90 86 90 91 93 90 Specialty Prod 75 73 74 79 81 76 Convnt hrs 94 81 81 90 95 88 Conv Location 85 85 90 82 84 85 No wait Chkout 76 63 57 83 85 73 Attractive store 82 77 86 85 85 83 Mean 82 77 81 86 86 82

Questions What is the relationship among stores? Which stores seem to be similar? Are there specific attributes on which some stores are rated better than on others? Are there groups of stores that are perceived similarly?

Removing Attribute Effect [Column (Store) effect] STORES Statement A B C D E Mean Friendly staff 5 7 9 7 5 7 Knowledgeable staff 0 4 7 3 0 3 Helpful staff -4 -3 -4 -3 -2 -3 Variety of Items 0 1 4 -1 -3 1 Satisfies Cust 5 3 8 6 8 6 Meets cust Expec -6 -6 -4 -3 -5 -4 Deli v on promises -2 -1 1 4 1 1 Meets needs -1 -2 2 2 2 1 Low credit int -10 -12 -15 -11 -7 -11 Everyday Shopng 8 9 9 5 7 8 Specialty Prod -7 -4 -7 -7 -5 -6 Convnt hrs 12 4 0 4 9 6 Conv Location 3 8 9 -4 -2 3 No wait Chkout -6 -14 -24 -3 -1 9 Attractive store 0 0 5 -1 -1 1 Mean 0 0 0 0 0 0

Row Attribute and Column (Store) effect Removed STORES Statement A B C D E Mean Friendly staff -2 0 2 0 -2 0 Knowledgeable staff -3 1 4 0 -3 0 Helpful staff -1 0 -1 0 1 0 Variety of Items -1 0 3 -2 -4 0 Satisfies Cust -1 -3 2 0 2 0 Meets cust Expec -2 -2 0 1 -1 0 Deli v on promises -3 -2 0 3 0 0 Meets needs -2 -3 1 1 1 0 Low credit int 1 -1 -4 0 4 0 Everyday Shopng 0 1 1 -3 -1 0 Specialty Prod -1 2 -1 -1 1 0 Convnt hrs 6 -2 -6 -2 3 0 Conv Location 0 5 6 -7 -5 0 No wait Chkout 3 -5 -15 6 8 0 Attractive store -1 -1 4 -2 -2 0 Mean 0 0 0 0 0 0

Store, Attribute and Random Noise effect Removed STORES Statement A B C D E Mean Friendly staff - - - - - 0 Knowledgeable staff - - 4 - - 0 Helpful staff - - - - - 0 Variety of Items - - - - -4 0 Satisfies Cust - - - - - 0 Meets cust Expec - - - - - 0 Deli v on promises - - - - - 0 Meets needs - - - - - 0 Low credit int - - -4 - 4 0 Everyday Shopng - - - - - 0 Specialty Prod - - - - - 0 Convnt hrs 6 - -6 - - 0 Conv Location - 5 6 -7 -5 0 No wait Chkout - -5 -15 6 8 0 Attractive store - - 4 - - 0 Mean 0 0 0 0 0 0

Some Patterns Stand Out clearly Store A is considered to have convenient hours; store C is not considered to have convenient hours The attribute ‘No wait for check out’ clearly discriminates stores : store C is perceived inferior while stores B, D and E are perceived superior Store E is perceived to lack variety Store E offers lower interest rates; store C does not

Patterns not obvious initially STORES Statement A B C D E Mean Friendly staff 87 84 90 93 91 89 Knowledgeable staff 82 81 88 89 86 85 Helpful staff 78 74 77 83 84 79 Variety of Items 82 78 85 85 83 83 Satisfied Customers 87 80 89 92 94 88 Meets cust Expec 76 71 77 83 81 78 Deli v on promises 80 76 82 90 87 83 Meets needs 81 75 83 88 88 83 Low credit int 72 65 66 75 79 71 Everyday Shopng 90 86 90 91 93 90 Specialty Prod 75 73 74 79 81 76 Convnt hrs 94 81 81 90 95 88 Conv Location 85 85 90 82 84 85 No wait Chkout 76 63 57 83 85 73 Attractive store 82 77 86 85 85 83 Mean 82 77 81 86 86 82

Procedure explained is known as ‘row and column centering’ The analysis can be made more sophisticated by standardizing the deviations

Learning so far There are hidden structures in data We can uncover such hidden structures by eliminating the variables that prevent us from seeing such patterns Such variables could be column effects, row effects, variability effects or some other effects Bi-variate analysis may not always be the best way to uncover structures Analysis to identify the structure can be carried out in more than one way

How are techniques classified? Two basic types of Analysis Structural and Functional

Structural Methods Which variables are similar? Which consumers can be grouped together for marketing purposes? Questions: Can we group consumers who are similar with regard to benefits they expect from a product? Can we identify variables that elicit a similar response from consumers?

Functional Methods Such techniques try to explain a variable in terms of other variables: What variables determine sales? What demographic characteristics predict a voter’s intention? Questions: Can we relate a consumer’s ratings of a product with his or her purchase intentions?

Commonly used Multivariate Techniques Structural Techniques Functional Techniques Factor Analysis Regression Analysis Cluster Analysis Covariance Analysis Perceptual mapping Discriminant Analysis Correspondence analysis Classification tree analysis

Primary Scales of Measurement Summary of Information contained Scale Difference Direction Rel Mag Abs. Mag Nominal Yes No No No Ordinal Yes Yes No No Interval Yes Yes Yes No Ratio Yes Yes Yes Yes

Techniques for Bivariate Analysis Independent Variable Dependant Variable Variable Nominal Interval/Ratio Nominal Chi Sq Discriminant Interval/Ratio ANOVA Reg/Cor

Understanding data

More Related Content

Similar to Understanding data (20)

More from Kinshook Chaturvedi (20)

Recently uploaded (20)

Understanding data