📘 Stata Lecture Notes – Preparing for Lab 3

📘 Stata Lecture Notes – Preparing for Lab 3#

Session Length: ~4 hours Goal: Students should confidently use if/else, create binary variables, write simple programs, loop over variable lists, and export regression tables.

🔹 1. Conditional Logic + Chi-Square#

💡 Stata can be taught to speak in complete English sentences — when you know how to control the logic.

Concepts Covered:

  • tabulate var1 var2, chi2

  • scalar, return list

  • display, if, else

Example:#

use transplants.dta, clear

tabulate rec_work gender, chi2

scalar p = r(p)

if p < 0.05 {
    display "Question 1: Working for income is associated with gender, p = " %5.3f p
}
else {
    display "Question 1: Working for income is NOT associated with gender, p = " %5.3f p
}

✅ Skills Gained: How to retrieve and use the p-value from a test using scalar and if/else.

🔹 2. Creating Clean Binary Variables#

🧠 Binary variables are the currency of logistic regression. You must be able to define your own currency.

Goal: Translate real-world conditions into clean 0/1 variables#

gen hypertensive = (dx == 4)

gen college = ///
    inlist(educ, 4, 5, 6)  // any college, associates, bachelors, grad

gen male = (gender == 1)   // assuming 1=male, 2=female

✳️ Tip: Use label define and label values to keep your binary variables readable.

🔹 3. Logistic Regression and Interpreting Odds Ratios#

logistic hypertensive age bmi college male

Then get readable odds ratios:

estimates store model1
esttab model1, eform

🔁 4. Writing a program to Display Odds Ratios in a Clean Format#

🔄 You don’t want to copy/paste regression output. Make Stata do it for you.

table_loop — Basic Loop Over Fixed Variables#

program define table_loop
    foreach var in age bmi college male {
        quietly logistic hypertensive `var'
        local or = exp(_b[`var'])
        local lb = exp(_b[`var'] - 1.96 * _se[`var'])
        local ub = exp(_b[`var'] + 1.96 * _se[`var'])
        display "`var'" _col(15) %4.2f `or' " (" %4.2f `lb' " - " %4.2f `ub' ")"
    }
end

table_loop

✅ Skills Gained: Using coefficient and standard error to build a confidence interval manually.

🧰 5. Generalizing the Program with syntax varlist#

🔧 Now that they’ve seen it hardcoded, time to teach them how to generalize it with syntax varlist.

table_varlist – Flexible Version#

program define table_varlist
    syntax varlist
    display "Regression table (Question 3)"
    foreach var of varlist `varlist' {
        quietly logistic hypertensive `var'
        local or = exp(_b[`var'])
        local lb = exp(_b[`var'] - 1.96 * _se[`var'])
        local ub = exp(_b[`var'] + 1.96 * _se[`var'])
        display "`var'" _col(15) %4.2f `or' " (" %4.2f `lb' " - " %4.2f `ub' ")"
    }
end

table_varlist age bmi college male

📊 6. Percentiles, Macros, and Variable Creation#

📈 How do you make a variable that depends on the 75th percentile of another?

summarize rec_hgt_cm, detail
scalar p75 = r(p75)
gen tall = (rec_hgt_cm >= p75)
label variable tall "Height ≥ 75th percentile"

Other variables:

gen white = (race == 1) // assuming 1 = white
gen age_10y = age / 10
label variable age_10y "Age (per 10y)"

📤 7. Exporting a Clean Regression Table to Excel#

📤 Professional tables come from putexcel.

logistic tall age_10y rec_wgt_kg gender white

putexcel set lab3_output.xlsx, replace
putexcel A1 = ("Variable") B1 = ("OR (95% CI)")

foreach var in age_10y rec_wgt_kg gender white {
    local or = exp(_b[`var'])
    local lb = exp(_b[`var'] - 1.96 * _se[`var'])
    local ub = exp(_b[`var'] + 1.96 * _se[`var'])
    local ci = string(`or', "%4.2f") + " (" + ///
                string(`lb', "%4.2f") + "-" + ///
                string(`ub', "%4.2f") + ")"
    putexcel A`=_n+1' = "`var'" B`=_n+1' = "`ci'"
}

🧠 End of Lecture: Recap + Stretch#

By now, they can:

  • Use conditional logic to control output

  • Define 0/1 variables cleanly

  • Run logistic models and interpret ORs

  • Write flexible programs that loop

  • Export clean tables to Excel

🧪 Stretch challenge (optional): Try making table_varlist also accept the outcome variable as a parameter.

Contents