- [email](https://mail.google.com/mail/u/0/#inbox/FMfcgzGkZZnHptrwzZWqjpBpJrtSRHph)
- N: 200-300 in India
- 6 weeks
- 40 unique headlines (20 true, 20 false)
- pseudorandom order (determined by us)
- post-survey (100 headlines)
- 40 exposed headlines (seen during 6 weeks)
- 60 new
- completely random order (can be determined by taskus)
- questions: vaccine intentions, COVID attitudes
- [ ] condition assignment
- control vs treatment (between-subjects)
- amusing/interesting vs. accurate?
- balance assignment as much as possible based on experience/time at taskus, gender, education, etc.
# Counterbalancing
```r
reps <- 5:1
n_true <- 4
n_false <- 4
dt1 <- data.table(reps, n_true, n_false)
dt1[, total_exposures := reps * n_true + reps * n_false]
dt1[, sum(total_exposures)] # total exposures
unique_headlines_n <- dt1[, sum(n_true) + sum(n_false)] # total headlines required
> unique_headlines_n # 20 false, 20 true
[1] 40
> dt1
reps n_true n_false total_exposures
1: 5 4 4 40 # reps * n_true + reps * n_false
2: 4 4 4 32
3: 3 4 4 24
4: 2 4 4 16
5: 1 4 4 8
> dt1[, sum(total_exposures)] # total exposures
[1] 120
```