Ch3 re write - combining research design and randomization

code/replicability.do → code/reproducible-randomization.do

@@ -19,5 +19,5 @@
     gen check3 = rnormal()  // Create a third random number after resetting seed
 
 * Visualize randomization results. See how check1 and check3 are identical,
-* but check2 is random relative check1 and check3
+* but check2 is random relative to check1 and check3
     graph matrix check1 check2 check3 , half

code/simple-multi-arm-randomization.do

@@ -1,20 +1,21 @@
-* Set up reproducbilitiy - VERSIONING, SORTING and SEEDING
+* Set up reproducible randmomization - VERSIONING, SORTING and SEEDING
     ieboilstart , v(13.1)    // Version
     `r(version)'             // Version
     sysuse bpwide.dta, clear // Load data
     isid patient, sort       // Sort
     set seed 654697          // Seed - drawn using https://bit.ly/stata-random
 
-* Generate a random number and use it to sort the observation. Then
-* the order the observations are sorted in is random.
+* Generate a random number and use it to sort the observation.
+* Then the order the observations are sorted in is random.
     gen treatment_rand = rnormal() // Generate a random number
     sort treatment_rand            // Sort based on the random number
 
-* See simple-sample.do example for an explanation of "(_n <= _N * X)". The code
-* below randomly selects one third of the observations into group 0, one third into group 1 and
-* one third into group 2. Typically 0 represents the control group and 1 and
-* 2 represents two treatment arms
-    generate treatment = 0                       // Set all observations to 0
+* See simple-sample.do example for an explanation of "(_n <= _N * X)".
+* The code below randomly selects one third of the observations into group 0,
+* one third into group 1 and one third into group 2.
+* Typically 0 represents the control group
+* and 1 and 2 represents the two treatment arms
+    generate treatment = 0                       // Set all observations to 0 (control)
     replace  treatment = 1 if (_n <= _N * (2/3)) // Set only the first two thirds to 1
     replace  treatment = 2 if (_n <= _N * (1/3)) // Set only the first third to 2
 

code/simple-uniform-probability-sampling.do

@@ -0,0 +1,26 @@
+* Set up reproducible randmomization - VERSIONING, SORTING and SEEDING
+    ieboilstart , v(13.1)    // Version
+    `r(version)'             // Version
+    sysuse bpwide.dta, clear // Load data
+    isid patient, sort       // Sort
+    set seed 215597          // Seed - drawn using https://bit.ly/stata-random
+
+* Generate a random number and use it to sort the observations.
+* Then the order the observations are sorted in is random.
+    gen sample_rand = rnormal() // Generate a random number
+    sort sample_rand            // Sort based on the random number
+
+* Use the sort order to sample 20% (0.20) of the observations.
+*_N in Stata is the number of observations in the active dataset,
+* and _n is the row number for each observation. The bpwide.dta has 120
+* observations and 120*0.20 = 24, so (_n <= _N * 0.20) is 1 for observations
+* with a row number equal to or less than 24, and 0 for all other
+* observations. Since the sort order is randomized, this means that we
+* have randomly sampled 20% of the observations.
+    gen sample = (_n <= _N * 0.20)
+
+* Restore the original sort order
+    isid patient, sort
+
+* Check your result
+    tab sample

manuscript.tex

@@ -60,16 +60,8 @@ \chapter{Chapter 2: Collaborating on code and data}
 %	CHAPTER 3
 %----------------------------------------------------------------------------------------
 
-\chapter{Chapter 3: Evaluating impact through research design}
-\label{ch:3}
 
-\input{chapters/research-design.tex}
-
-%----------------------------------------------------------------------------------------
-%	CHAPTER 4
-%----------------------------------------------------------------------------------------
-
-\chapter{Chapter 4: Sampling, randomization, and power}
+\chapter{Chapter 3: Establish a measurement framework}
 \label{ch:4}
 
 \input{chapters/sampling-randomization-power.tex}