Fitted separate DDMs to discordant and concordant trials. About half the number of trials for each model as a result.
# Model fits
Surprisingly good model fits.
## Discordant (`politically_concordant == 0`)
![[rating2_pred 4.png]]
![[rt_pred 4.png]]
![[rt0_pred 4.png]]
![[rt1_pred 4.png]]
## Concordant (`politically_concordant == 1`)
![[rating2_pred 5.png]]
![[rt_pred 5.png]]
![[rt0_pred 5.png]]
![[rt1_pred 5.png]]
# Parameters
## Drift
![[drift.png|1200]]
```r
> summaryh(lmer(drift ~ conditionEC * modelEC * platformEC + (1 + conditionEC * modelEC | study), d11))
boundary (singular) fit: see ?isSingular
term results
1: (Intercept) b = −0.31, SE = 0.06, t(6) = −5.29, p = .002, r = 0.90
2: conditionEC b = −0.02, SE = 0.05, t(6) = −0.44, p = .672, r = 0.17
3: modelEC b = 0.24, SE = 0.06, t(10) = 3.66, p = .005, r = 0.76 # drift rate is higher for concordant headlines
4: platformEC b = −0.09, SE = 0.12, t(6) = −0.81, p = .450, r = 0.30
5: conditionEC:modelEC b = 0.08, SE = 0.09, t(129) = 0.84, p = .404, r = 0.07
6: conditionEC:platformEC b = −0.006, SE = 0.11, t(6) = −0.06, p = .955, r = 0.02
7: modelEC:platformEC b = 0.05, SE = 0.13, t(10) = 0.38, p = .714, r = 0.12
8: conditionEC:modelEC:platformEC b = −0.25, SE = 0.19, t(129) = −1.33, p = .187, r = 0.12
```
## Boundary
![[bound.png|1200]]
```r
> summaryh(lmer(B ~ conditionEC * modelEC * platformEC + (1 + conditionEC | study), d11))
boundary (singular) fit: see ?isSingular
term results
1: (Intercept) b = 3.23, SE = 0.12, t(5) = 27.90, p < .001, r = 1.00
2: conditionEC b = 0.02, SE = 0.16, t(5) = 0.13, p = .899, r = 0.06
3: modelEC b = −0.36, SE = 0.12, t(2283) = −2.95, p = .003, r = 0.06 # boundary is reduced for concordant headlines
4: platformEC b = −0.42, SE = 0.23, t(5) = −1.82, p = .130, r = 0.64
5: conditionEC:modelEC b = −0.29, SE = 0.25, t(2283) = −1.16, p = .245, r = 0.02
6: conditionEC:platformEC b = −0.11, SE = 0.33, t(5) = −0.34, p = .748, r = 0.16
7: modelEC:platformEC b = −0.58, SE = 0.25, t(2283) = −2.35, p = .019, r = 0.05 # reduced boundary effect is stronger for MTurk
8: conditionEC:modelEC:platformEC b = 0.07, SE = 0.49, t(2283) = 0.14, p = .889, r = 0.003
```
## Starting point
![[bias 7.png|1200]]
```r
> summaryh(lmer(x0 ~ conditionEC * modelEC * platformEC + (1 + conditionEC | study), d11))
boundary (singular) fit: see ?isSingular
term results
1: (Intercept) b = −0.08, SE = 0.03, t(5) = −2.76, p = .038, r = 0.77
2: conditionEC b = −0.03, SE = 0.03, t(891) = −0.80, p = .424, r = 0.03
3: modelEC b = 0.06, SE = 0.03, t(2280) = 1.97, p = .050, r = 0.04 # increased starting point for concordant headlines
4: platformEC b = 0.02, SE = 0.06, t(5) = 0.31, p = .769, r = 0.13
5: conditionEC:modelEC b = 0.10, SE = 0.07, t(2280) = 1.46, p = .143, r = 0.03
6: conditionEC:platformEC b = 0.01, SE = 0.07, t(891) = 0.17, p = .865, r = 0.006
7: modelEC:platformEC b = 6e−04, SE = 0.07, t(2280) = 0.01, p = .992, r = 2e−04
8: conditionEC:modelEC:platformEC b = −0.07, SE = 0.13, t(2280) = −0.56, p = .572, r = 0.01
```
## Non-decision time
![[ndt 7.png|1200]]
```r
> summaryh(lmer(nondectime ~ conditionEC * modelEC * platformEC + (1 | study), d11))
boundary (singular) fit: see ?isSingular
term results
1: (Intercept) b = 3.74, SE = 0.07, t(2285) = 55.61, p < .001, r = 0.76
2: conditionEC b = −0.12, SE = 0.13, t(2285) = −0.92, p = .355, r = 0.02
3: modelEC b = 0.19, SE = 0.13, t(2285) = 1.40, p = .161, r = 0.03
4: platformEC b = 0.06, SE = 0.13, t(2285) = 0.46, p = .645, r = 0.01
5: conditionEC:modelEC b = 0.14, SE = 0.27, t(2285) = 0.51, p = .610, r = 0.01
6: conditionEC:platformEC b = 0.12, SE = 0.27, t(2285) = 0.46, p = .648, r = 0.01
7: modelEC:platformEC b = 0.46, SE = 0.27, t(2285) = 1.72, p = .085, r = 0.04 # close to significance
8: conditionEC:modelEC:platformEC b = −0.23, SE = 0.54, t(2285) = −0.44, p = .663, r = 0.009
```