Eid, S. (2010, October 04). Data Literacy 101. Presentation Presented at: Lehigh Valley Health Network, Allentown, PA.

Size: px

Start display at page:

Download "Eid, S. (2010, October 04). Data Literacy 101. Presentation Presented at: Lehigh Valley Health Network, Allentown, PA."

Georgiana Houston
5 years ago
Views:

Lehigh Valley Health Network LVHN Scholarly Works Department of Community Health and Health Studies Data Literacy 101. Sherrine Eid MPH Lehigh Valley Health Network, Sherrine.Eid@lvhn.

1 Lehigh Valley Health Network LVHN Scholarly Works Department of Community Health and Health Studies Data Literacy 101. Sherrine Eid MPH Lehigh Valley Health Network, Follow this and additional works at: community_health_and_health_studies Part of the Community Health and Preventive Medicine Commons, and the Health Services Research Commons Published In/Presented At Eid, S. (2010, October 04). Data Literacy 101. Presentation Presented at: Lehigh Valley Health Network, Allentown, PA. This Presentation is brought to you for free and open access by LVHN Scholarly Works. It has been accepted for inclusion in LVHN Scholarly Works by an authorized administrator. For more information, please contact

2 Data Literacy 101 Sherrine Eid, MPH Biostatistician 1

3 About this mini-course Introductions Syllabus Logistics 2

4 My level of expertise in data literacy is 1. I have absolutely no clue what you are talking about! 2. I ve worked on 1 or 2 research projects 3. I m a PI on several research projects 4. I design research projects for a living I have absolut... I ve worked on... I m a PI on se... I design resea

5 I m taking this workshop because 1. Strengthen research skills 2. Acquire research skills 3. Job requirement 10 4

6 Units Unit 1. Introduction to measurement Unit 2. The systematic inquiry process Unit 3. The two-by-two table and the chi-square test 5

7 Unit 1: Introduction to measurement 6

8 Unit 1 Pretest 7

9 Body weight is a variable. 1. Dichotomous 2. Polychotomous 3. Continuous 10 8

10 Categories of categorical variables should be exclusive AND mutually exhaustive. 1. True 2. False 10 True False 9

11 Which method is the more desirable in terms of Informativeness? Consider a survey measuring the ages of 100 children. Method 1 Which age group best represents the age in years of the patient? a. 0-3 b c Method 2 What is the age of the patient in years? 1. Method 1 2. Method 2 Method 1 Method

12 An investigator hypothesizes that the adult weight of a dog is higher when it has fewer littermates. The dog s weight would be 1. Independent variable 2. Dependent variable 10 11

13 An investigator hypothesizes that the adult weight of a dog is higher when it has fewer littermates. The number of littermates would be 1. Independent variable 2. Dependent variable 10 12

14 A group of healthcare professionals took a research course to improve their research skills. The course in this case would be the 1. Independent variable 2. Dependent variable 10 13

15 Variable Definition: Any attribute, phenomenon, or event that can have different values Sub-types Continuous Categorical Dichotomous Polychotomous Independent Dependent 14

16 Continuous variables Definition: Have a countless (or infinite) number of values along a continuum Example Give 100 people a questionnaire One question asks, What is your yearly income? Possible answers ($): 25,000; 25,550; 95,000; 109,000; etc. 15

17 Continuous Data Mean, standard deviation, or Median and interquartile range, where appropriate. Length of Stay Weight LDL 1, 2, 3, M 3 M 6 M 9 M 12 M 15 M Control Intervention 16

18 Categorical variables Definition: Have countable (or finite) number of values; contain distinct categories Examples Gender (male, female) Yearly income (low, medium, high) 17

Categorical Variables Dichotomous variable. Has 2 countable values, or distinct categories Example. Survey question asks, What is your gender?

19 Categorical Variables Dichotomous variable. Has 2 countable values, or distinct categories Example. Survey question asks, What is your gender? ( male, female) Polychotomous variable. Has 3 or more countable values, or distinct categories Survey question asks, What is your yearly income? ( low; medium; high) 18

20 Polychotomous Data Median and interquartile range. Age groups Ranges BIG steps , 18-24, 25-44, , 60-79, 80+yr Control Intervention 19

21 Nominal Data Proportion or rate with a confidence interval. Male/Female Control/Intervention >50/<50 YES/NO 43% 57% Female Male 20

22 Proper categories for categorical Balanced variables are: Not too few or too many categories Rule-of-thumb: a categorical variable should have 3 to 5 categories Comparable to standards Use Federal questionnaires (e.g., U.S. census) as guides 21

Every possible observation fits into a category Non-example Survey

23 Proper categories for categorical variables are: Exhaustive Definition. Every possible observation fits into a category Non-example Survey question In thinking about your weight, do you consider yourself to be: Overweight About the right weight What is missing here? 22

24 Proper categories for categorical variables are: Exhaustive Example In thinking about your weight, do you consider yourself to be: Overweight About the right weight Underweight 23

25 Proper categories for categorical variables are: Mutually-exclusive Definition. No observation can logically be placed in more than one category Non-Example Survey question among adults What is your age? 18 to 39 years 39 to 59 years 60 to 89 years 90 years or older What is wrong here? 24

26 Proper categories for categorical variables are: Mutually-exclusive Example Survey question among adults What is your age? 18 to 39 years 40 to 59 years 60 to 89 years 90 years or older 25

Informativeness Better to collect data using continuous variables than categorical variables Because continuous variables are more informative, efficient Example Means

27 Informativeness Better to collect data using continuous variables than categorical variables Because continuous variables are more informative, efficient Example Means your project/study will have more power Each person s years of age (continuous variable) Vs. classifying each person as younger or older, which is dichotomous variable 26

28 Scaling Down Even if you collect data using a continuous variable (e.g., years of age), you can scale it down to a polychotomous variable later, in the analysis phase If years of age =18-39, then age = young If years of age = 40-59, then age = middle-aged If years of age= 60+, then age = senior 27

29 Scaling Down You can also scale down a continuous variable to a dichotomous variable later, in the analysis phase If years of age =18-59, then age = young If years of age = 60+, then age = old 28

30 Scaling Up Although you can scale down You CANNOT scale up If you collect data using a dichotomous/polychotomous variable, you can t turn it into a continuous variable later in the analysis phase 29

31 Scaling Up: Example Survey question among adults What is your age? 18 to 39 years 40 to 59 years 60 to 89 years 90 years or older You can t turn this into a continuous variable; you don t know exact values within a category 30

32 Independent and dependent variables Independent variable Definition: Variable that influences another variable Related term. Exposure variable (E) Dependent variable Definition: The variable which is influenced by the independent variable Related term. Disease variable (D) 31

33 Independent and dependent variables: Examples 1. High exposure to lead paint is associated with mental retardation. Lead paint is independent variable Mental retardation is dependent variable 2. People with low literacy levels are more likely to be obese. Independent variable? Dependent variable? 32

34 Summary Variable sub-types Continuous Categorical Dichotomous Polychotomous Independent Dependent Proper categories for categorical variables Informativeness Scaling up Scaling down 33

35 Unit 1 Posttest 34

36 Gender is a variable. 1. Dichotomous 2. Polychotomous 3. Continuous 10 Dichotomous Polychotomous Continuous 35

37 Age in years is a variable. 1. Dichotomous 2. Polychotomous 3. Continuous 10 Dichotomous Polychotomous Continuous 36

38 Gross annual income in dollars is a variable. 1. Dichotomous 2. Polychotomous 3. Continuous 10 Dichotomous Polychotomous Continuous 37

39 Gross annual income in dollars (below average, average, above average) is a variable. 1. Dichotomous 2. Polychotomous 3. Continuous 10 Dichotomous Polychotomous Continuous 38

40 Education (high school degree, no high school degree) is a variable. 1. Dichotomous 2. Polychotomous 3. Continuous 10 Dichotomous Polychotomous Continuous 39

41 Categories of categorical variables should be exclusive AND mutually exhaustive. 1. True 2. False 10 True False 40

42 Which method is the more desirable in terms of Informativeness? Consider a survey measuring the ages of 100 children. Method 1 Which age group best represents the age in years of the patient? a. 0-3 b c Method 2 What is the age of the patient in years? 1. Method 1 2. Method 2 Method 1 Method

43 If you used Method 2, can you take the actual ages of the 100 patients, and rescale them into the ordered categories of Method 1? Consider a survey measuring the ages of 100 children. Method 1 Which age group best represents the age in years of the patient? a. 0-3 b c Method 2 What is the age of the patient in years? 1. Yes 2. No Yes No 10 42

44 If you used Method 1, can you take the ordered categories and rescale them into the actual ages of the 100 patients (Method 2)? Consider a survey measuring the ages of 100 children. Method 1 Which age group best represents the age in years of the patient? a. 0-3 b c Method 2 What is the age of the patient in years? 1. Yes 2. No Yes No 10 43

A group of students were given a short course in speed-reading. The instructor was curious if a monetary incentive would influence performance on a reading test taken at the end of the course.

45 A group of students were given a short course in speed-reading. The instructor was curious if a monetary incentive would influence performance on a reading test taken at the end of the course. Half of the students were offered $5 for obtaining a certain level of performance on the test, the other half were not offered money. The monetary incentive in this case would be the 1. Independent variable 2. Dependent variable 10 44

46 An investigator hypothesizes that the adult weight of a dog is higher when it has fewer littermates. The dog s weight would be 1. Independent variable 2. Dependent variable 10 45

47 Batches of seeds are soaked in, 5%, 1, and 2 salt solutions for concentrations and germination is counted for each batch. Germination would be 1. Independent variable 2. Dependent variable 10 46

48 Batches of seeds are soaked in, 5%, 1, and 2 salt solutions for concentrations and germination is counted for each batch. Is the independent variable continuous, dichotomous, or polychotomous? 1. Continuous 2. Dichotomous 3. Polychotomous 10 47

49 Batches of seeds are soaked in salt solutions for different concentrations and germination is counted for each batch. Is the dependent variable continuous, dichotomous, or polychotomous? 1. Continuous 2. Dichotomous 3. Polychotomous 10 48

50 Group Exercise Lab 1 49

51 Unit 2: The Systematic Inquiry Process 50

52 Unit 2 Pretest 51

53 Does smoking cigarettes lead to lung cancer? Is a focus question. 1. True 2. False 10 True False 52

54 Research question=focus question. 1. True 2. False 10 True False 53

55 Concept maps can be used to 1. Can be used to find your travel destination 2. Are useless 3. Can be used to refine you research question 4. Can be used to determine IRB approval 10 54

56 Systematic Inquiry Definition. A careful, deliberate effort to solve a problem, to investigate something inadequately known or understood. 55

57 Systematic Inquiry: Process 1. Identify the problem 2. Review the literature 3. Specify your focus question 4. Design your project/study 5. Collect data 6. Analyze and interpret data 7. Report findings 56

58 Systematic Inquiry: Emphasis in this course 1. Identify the problem 2. Review the literature 3. Specify your focus question 4. Design your project/study 5. Collect data 6. Analyze and interpret data 7. Report findings 57

59 Focus Question Definition. Question about a topic that you do not know or fully understand Features. Two parts Independent (or exposure) variable Dependent (or disease) variable Examples Does smoking cigarettes lead to lung cancer? Does participation in a computer training program increase the likelihood of subsequent employment? 58

60 Focus Question Related terms Research question Evaluation question 59

61 Where do focus questions come from? Problem presenting itself in the work setting Popular controversy Issue receiving a lot of coverage in newspapers or academic journals You or a family member has suffered from a disease condition 60

62 Characteristics of good focus questions Important So what? Who cares? Answerable Not What is the meaning of life? Feasible Do the investigators have the necessary skill, background, resources? Grounded in broader theory (social, biological, behavioral, statistical) Can be captured in Concept Map 61

63 Concept Map Diagram showing relationships among concepts Enhances metacognition Details the entire structure of an idea Communicates complex ideas and arguments Used in brainstorming 62

64 63

65 Unit 2 Posttest 64

66 Does participation in a computer training program increase the likelihood of subsequent employment? Is a focus 1. True 2. False question. 10 True False 65

67 Evaluation question=focus question. 1. True 2. False 10 True False 66

68 Concept maps 1. Can be used to find your travel destination 2. Can t be used 3. Can be used to refine you research question 4. Can be used to determine IRB approval 10 67

69 Unit 3: The two-by-two table and the chi-square test 68

70 Unit 3 Pretest 69

71 A 2x2 table 1. Is not used in research 2. Tests IV and DV associations 3. Tests the mean difference between two groups 10 70

72 Noise and sound can be found in healthcare research. 1. True 2. False 10 True False 71

73 The large ratio from a chi square means 1. The exposure and disease are related 2. The exposure and disease are not related 3. The exposure and disease are the same 4. The exposure and the disease are different 10 The exposure a... The exposure a... The exposure a... The exposure a... 72

74 Transition One wants a focus question, which asks whether the independent variable relates to the dependent variable BUT, how do we answer the focus question? How will we know that the 2 variables relate to each other? Consider radio broadcast engineering. 73

Radio broadcast engineering (RBE) You re driving your car and searching for particular channel on the radio As you go from channel to channel, you hear static on

75 Radio broadcast engineering (RBE) You re driving your car and searching for particular channel on the radio As you go from channel to channel, you hear static on almost every frequency Background noise (N) drowning out music signal (S); N > S or S < N E.g., S = 10 value, N = 100 value S/N ratio =10/100 = 1/10; considered small 74

76 Radio broadcast engineering (RBE) You continue to try different channels Eventually, reach a point where S is strong, and N minimal; stay there S > N E.g., S = 100 value, N= 10 value S/N ratio = 100/10 = 10 Considered large 75

77 Radio broadcast engineering In health projects, relationship between independent and dependent variables has S and N too!!! So just like radio scenario, situations in health where sometimes S<N and other times S>N We want S>N or large S/N ratio 76

78 2x2 Table Definition. Tabular cross-classification of data, whereby the independent variable (or E) is displayed horizontally (in rows); and the dependent variable (or D) is displayed vertically (in columns) E and D variables must both be dichotomous variables Significance Can calculate S/N Most widely-used data display 77

79 Standard 2 x 2 Table D (+) Col. 1 D (-) Col. 2 Total E (+) Row 1 E (-) Row 2 a b a + b c d c + d Total a + c b + d a + b + c + d = n (overall sample) 78

Example Cell a represents the number of smokers with CA Cell b represents the number of smokers without CA Cell c represents the number of non-smokers with CA Cell d represents the

80 Example Cell a represents the number of smokers with CA Cell b represents the number of smokers without CA Cell c represents the number of non-smokers with CA Cell d represents the number of non-smokers without CA Smoking (E+) Non smoking (E-) Lung CA (D+) 76 (a) 12 (c) Total 88 No lung CA (D-) 20 (b) 98 (d) 118 Total 96 (a+b) 110 (c+d) 206 (a+c) (b+d) (n) 79

Example (continued) Cell a + b represents the number of smokers Lung CA (D+) No lung CA (D-) Total Cell c + d represents the number of non-smokers Cell a + c represents the number with CA Smoking

81 Example (continued) Cell a + b represents the number of smokers Lung CA (D+) No lung CA (D-) Total Cell c + d represents the number of non-smokers Cell a + c represents the number with CA Smoking (E+) 76 (a) 20 (b) 96 (a+b) Cell b + d represents the number without CA Cell n represents the total number of study subjects Non smoking (E-) 12 (c) Total (d) (c+d) 206 (a+c) (b+d) (n) 80

82 Example (continued) We to calculate S/N ratio Lung CA (D+) No lung CA (D-) Total Called chi-square test statistic Smoking (E+) 76 (a) 20 (b) 96 (a+b) Formula = n(ad-bc) 2 (a+c)(b+d)(a+b)(c+d) Plug in numbers Non smoking (E-) 12 (c) Total 88 (a+c) 98 (d) 118 (b+d) 110 (c+d) 206 (n) 81

83 Example (continued) Formula = Lung CA No lung CA Total n(ad-bc) 2 (D+) (D-) (a+c)(b+d)(a+b)(c+d) Don t have to perform computations yourself Just know that numerator = S; and denominator = N Smoking (E+) Non smoking (E-) 76 (a) 12 (c) Total 88 (a+c) 20 (b) 98 (d) 118 (b+d) 96 (a+b) 110 (c+d) 206 (n) 82

84 Example (continued) Computer performs computations Lung CA (D+) No lung CA (D-) Total S/N ratio = 58.3 Considered large ratio Conclusion. Smoking is statistically associated with lung CA RULE: ratio considered large if >3.84 Smoking (E+) Non smoking (E-) 76 (a) 12 (c) Total 88 (a+c) 20 (b) 98 (d) 118 (b+d) 96 (a+b) 110 (c+d) 206 (n) 83

85 Example (continued) Eyeball the data to make sure S/N ratio makes sense Among smokers (n=96), most (n=76) have CA Among nonsmokers (n=110), most (n=98) don t have CA Smoking (E+) Non smoking (E-) Lung CA (D+) 76 (a) 12 (c) Total 88 (a+c) No lung CA (D-) 20 (b) 98 (d) 118 (b+d) Total 96 (a+b) 110 (c+d) 206 (n) 84

OpenEpi Software Web-based software for public use Developed at Emory University School of Public Health www.

86 OpenEpi Software Web-based software for public use Developed at Emory University School of Public Health Select Counts Select Two by Two Table Plug in your numbers Be careful of other statistical tests offered 85

Quality improvement/intervention example Focus question, Does full participation in a computer training program influence subsequent employment? Chi-square = 12.5 >3.

87 Quality improvement/intervention example Focus question, Does full participation in a computer training program influence subsequent employment? Chi-square = 12.5 >3.84 therefore there is a statistical association Eyeballing the data backs up statistics Full participation (E+) Partial participation (E-) Employed (D+) 30 (a) 70 (c) Total 100 (a+c) Not Employed (D-) 10 (b) 90 (d) 100 (b+d) Total 40 (a+b) 160 (c+d) 200 (n) 86

88 Unit 3 Posttest 87

89 A 2x2 table 1. Is not used in research 2. Tests E and D associations 3. Tests the mean difference between two groups 10 88

90 Noise and sound can be found in healthcare research. 1. True 2. False 10 True False 89

91 If N>S, there is an association between E and D. 1. True 2. False 10 True False 90

The large ratio from a chi square means 10 1.

92 The large ratio from a chi square means The E and D are related 2. The E and D are not related 3. The E and D are the same 4. The E and the D are different The E and D are related The E and D are not r... The E and D are the... The E and the D are di... 91

The ratio from a chi square, the numerator measures 10 1. How to get from point A to point B 2. The strength of the relationship 3.

93 The ratio from a chi square, the numerator measures How to get from point A to point B 2. The strength of the relationship 3. The positive predictive value 4. The sound in the N/S paradigm 5. The noise in the N/S paradigm How to get fro... The strength o... The positive p... The sound in t... The noise in t... 92

94 Contact Information:

95 Resources: 94

96 GROUP ACTIVITY 95

97 Course Evaluation 96

Assessment Schedule 2015 Mathematics and Statistics (Statistics): Evaluate statistically based reports (91584)

Assessment Schedule 2015 Mathematics and Statistics (Statistics): Evaluate statistically based reports (91584) NCEA Level 3 Mathematics and Statistics (Statistics) (9584) 205 page of 7 Assessment Schedule 205 Mathematics and Statistics (Statistics): Evaluate statistically based reports (9584) Evidence Statement