2016-03-25 - example: harmonizing smoking

[andkov]

Notes during meeting with @smhofer

GOAL

create two binary variables:

• 1. Are you a current smoker? ( 0 - no , 1 - yes)
• 1. Have you ever smoked? ( 0 - no , 1 - yes)

     Random Notes

• in SATSA, two items ( https://www.maelstrom-research.org/mica/variable/satsa_q3%3AGSMOKNOW%3AStudy and https://www.maelstrom-research.org/mica/variable/satsa_q3%3AGEVRSMK%3AStudy) are tapping into the same "smoke_now". Do data verification to confirm that the two variables are telling the same thing. (prefer more information).
• in SHARE the question about smoke_history read " Every smoked for at least a year?". Other studies may ask about the smoking history in less specific term.

  Notes on Harmonization :

• ensure "sex" is coded the same
• bring age on the same metric
• make assumption about the year of testing (look up the year when the wave was measured)
• center age at the same number (65)

  marital status

• 0 - not married, single, divorced, widowed
• 1 - married
• (Or the common least complex categorization)

  Education

• consider a couple of categorizations

• we keep it as simple as possible

  one possible way:

• 0 - less then high-school
• 1 - high-school most
• 2 - college
• 3 - college plus

  BMI

• underweight (BMI less than 18.5)
• normal weight (BMI between 18.5 & 24.9)
• overweight (BMI between 25.0 & 29.9)
• obese (BMI 30.0 and above)

see more at http://www.whathealth.com/bmi/formula.html, bring on the same scale with height and weight. May need to computes.

Alcohol

• 0 : abstainer
• 2 : drinker

  Self-rated health

• play around, see what come up, discuss in meeting next time.

Suggested:

• 0 : poor/very poor
• 1 : goog/ very good

  Physical activity

• put off for the next week, because this presents the most challange

  Analytic goals

• Logistic regression
• report risk/odds relative to "better" conditions
• changes and risks relative to the reference group per unit changes
• the "0" must be good, indicating a "better", salubrious choice among, whereas "1" should indicate a "less healthy" choice.

[andkov]

Plan for the reproducible report(s):

• Section 1: Read in each of five data sets
• Section 2: Relabel and transform variables (organized by data set; as discussed yesterday)
• Section 3: Combine into single data set (include study level dummy variables)
• Section 4: Estimate models (ever smoked as primary outcome to start; logistic regression)
• Section 5: Table results and compute odds ratio for covariates

Here is a good ref for the analysyis in r http://www.ats.ucla.edu/stat/r/dae/logit.htm