Chapter 28 [fMRI] Kragel2018 ~ singletrial

What is the purpose of this notebook?

Here, I model NPS dot products as a function of cue, stimulus intensity and expectation ratings.
One of the findings is that low cues lead to higher Cognitive Control dotproducts in the medium intensity group, and that this effect becomes non-significant across sessions.
03/23/2023: For now, I’m grabbing participants that have complete data, i.e. 18 runs, all three sessions. N = 65

## [1] "/Users/h/Dropbox (Dartmouth College)/projects_dropbox/social_influence_analysis/analysis/fmri/nilearn/signature_extract/signature-Kragel18CogControl_sub-all_runtype-pvc_event-stimulus.tsv"

pvc.sub$trial_ind <- pvc.sub$trial_index -1 
pvc.sub$sub <- sprintf("sub-%04d", pvc.sub$src_subject_id)
pvc.sub$ses <- sprintf("ses-%02d", pvc.sub$session_id)
pvc.sub$run <- sprintf("run-%02d", pvc.sub$param_run_num)
pvc.sub$runtype <- sprintf("runtype-%s", pvc.sub$param_task_name)
pvc.sub$trial <- sprintf("trial-%03d", pvc.sub$trial_ind)
pvc.sub[c('cue', 'DEPc')]  <- str_split_fixed(pvc.sub$param_cue_type , '_', 2)
pvc.sub$cuetype <- sprintf("cuetype-%s", pvc.sub$cue)
pvc.sub[c('stimintensity', 'DEP')]  <- str_split_fixed(pvc.sub$param_stimulus_type , '_', 2)

# merge
pvc.beh <- pvc.sub[,c("sub", "ses", "run", "runtype", "trial", "cuetype", "stimintensity","event02_expect_angle", "event04_actual_angle")]
df_merge <- merge(pvc, pvc.beh, 
                  by.x = c("sub", "ses", "run", "runtype", "trial", "cuetype", "stimintensity"),
                  by.y = c("sub", "ses", "run", "runtype", "trial", "cuetype", "stimintensity")
                  )

28.1 Kragel2018 ~ 3 task * 3 stimulus_intensity

Q. What does the Cognitive control pattern look like for the three tasks?

Table 28.1: Contrast weights
	pain	vicarious	cognitive
task_V_gt_P	-0.50	0.50	0.00
task_C_gt_PV	-0.34	-0.34	0.66

model.alltask <- lmer(Kragel18CogControl ~ task_V_gt_P*stim_factor + task_C_gt_PV*stim_factor + (task|sub), data = pvc)
sjPlot::tab_model(model.alltask)

	Kragel18CogControl
Predictors	Estimates	CI	p
(Intercept)	-0.03	-0.04 – -0.03	<0.001
task V gt P	0.03	0.02 – 0.04	<0.001
stim factor [low]	0.00	-0.00 – 0.00	0.720
stim factor [med]	0.00	-0.00 – 0.00	0.810
task C gt PV	0.02	0.01 – 0.03	<0.001
task V gt P * stim factor [low]	0.01	0.01 – 0.02	0.001
task V gt P * stim factor [med]	0.00	-0.00 – 0.01	0.431
stim factor [low] * task C gt PV	0.01	0.01 – 0.02	<0.001
stim factor [med] * task C gt PV	0.01	-0.00 – 0.01	0.101
Random Effects
σ²	0.01
τ₀₀ _sub	0.00
τ₁₁ _sub.taskpain	0.00
τ₁₁ _{sub.taskvicarious}	0.00
ρ₀₁	-0.69
	-0.66
ICC	0.19
N _sub	111
Observations	19428
Marginal R² / Conditional R²	0.043 / 0.223

eta-squared

Parameter	Eta2_partial	CI	CI_low	CI_high
task_V_gt_P	0.3009101	0.95	0.1853008	1
stim_factor	0.0000070	0.95	0.0000000	1
task_C_gt_PV	0.3141347	0.95	0.1981308	1
task_V_gt_P:stim_factor	0.0006843	0.95	0.0001644	1
stim_factor:task_C_gt_PV	0.0010896	0.95	0.0004037	1

Cohen’s d

	t	df	d
task_V_gt_P	5.3894497	136.0766	0.9240228
stim_factorlow	0.3579918	19085.1501	0.0051827
stim_factormed	0.2398900	19085.1501	0.0034729
task_C_gt_PV	4.8174290	153.0070	0.7789143
task_V_gt_P:stim_factorlow	3.4494916	19085.1501	0.0499387
task_V_gt_P:stim_factormed	0.7878756	19085.1501	0.0114062
stim_factorlow:task_C_gt_PV	4.5069919	19085.1501	0.0652483
stim_factormed:task_C_gt_PV	1.6383590	19085.1501	0.0237187

28.2 Kragel2018 ~ 2 cue x 3 stimulus_intensity

Q. Within cognitive task, does stimulus intenisty level and cue level significantly predict Kragel2018 dotproducts?

28.2.1 Linear model ver 1: Kragel2018 ~ stimulus_intenisy * cue

model.cogstim <- lmer(Kragel18CogControl ~ stim_con_linear + stim_con_quad +(stim_con_linear + stim_con_quad |sub), data = data_screen)

## boundary (singular) fit: see help('isSingular')

sjPlot::tab_model(model.cogstim)

	Kragel18CogControl
Predictors	Estimates	CI	p
(Intercept)	-0.01	-0.02 – 0.00	0.214
stim con linear	-0.01	-0.02 – -0.00	0.002
stim con quad	-0.00	-0.01 – 0.00	0.266
Random Effects
σ²	0.01
τ₀₀ _sub	0.00
τ₁₁ _{sub.stim_con_linear}	0.00
τ₁₁ _{sub.stim_con_quad}	0.00
ρ₀₁	0.87
	0.93
N _sub	71
Observations	2743
Marginal R² / Conditional R²	0.004 / NA

Linear model ver 1: eta-squared

Parameter	Eta2_partial	CI	CI_low	CI_high
stim_con_linear	0.1249270	0.95	0.0282714	1
stim_con_quad	0.0030758	0.95	0.0000000	1

Linear model ver 1: Cohen’s d: Cognitive control stimulus intensity = 0.72

## boundary (singular) fit: see help('isSingular')

	t	df	d
stim_con_linear	-3.123369	68.33365	-0.7556766
stim_con_quad	-1.112014	400.79623	-0.1110909

28.2.2 Linear model ver 2: Kragel2018 ~ stimulus_intenisy * cue

## boundary (singular) fit: see help('isSingular')

	Kragel18CogControl
Predictors	Estimates	CI	p
(Intercept)	-0.01	-0.02 – 0.00	0.192
CUE high gt low	-0.00	-0.01 – 0.00	0.136
stim con linear	-0.01	-0.02 – -0.00	0.002
stim con quad	-0.00	-0.01 – 0.00	0.317
CUE high gt low * stim con linear	0.00	-0.01 – 0.02	0.744
CUE high gt low * stim con quad	-0.01	-0.02 – 0.01	0.320
Random Effects
σ²	0.01
τ₀₀ _sub	0.00
τ₁₁ _{sub.CUE_high_gt_low}	0.00
ρ₀₁ _sub	1.00
N _sub	71
Observations	2743
Marginal R² / Conditional R²	0.005 / NA

Linear model ver 2: eta-squared

Parameter	Eta2_partial	CI	CI_low	CI_high
CUE_high_gt_low	0.0074001	0.95	0.0000000	1
stim_con_linear	0.0035597	0.95	0.0007787	1
stim_con_quad	0.0003745	0.95	0.0000000	1
CUE_high_gt_low:stim_con_linear	0.0000399	0.95	0.0000000	1
CUE_high_gt_low:stim_con_quad	0.0003703	0.95	0.0000000	1

Linear model ver 2: Cohen’s d: CognitiveControl stimulus_intensity d = 0.73, cue d = 0.069

	t	df	d
CUE_high_gt_low	-1.4919521	298.5707	-0.1726877
stim_con_linear	-3.0896597	2672.1084	-0.1195401
stim_con_quad	-1.0005545	2672.3127	-0.0387104
CUE_high_gt_low:stim_con_linear	0.3265372	2674.7348	0.0126276
CUE_high_gt_low:stim_con_quad	-0.9950382	2672.7716	-0.0384936

Lineplots

28.2.3 2 Cue x 3 Stimulus intensity x Expectation rating

model.cog3factor <- lmer(Kragel18CogControl ~ CUE_high_gt_low*stim_con_linear*event02_expect_angle + CUE_high_gt_low*stim_con_quad*event02_expect_angle + (CUE_high_gt_low|sub), data = data_screen)

## boundary (singular) fit: see help('isSingular')

sjPlot::tab_model(model.cog3factor)

	Kragel18CogControl
Predictors	Estimates	CI	p
(Intercept)	-0.01	-0.02 – 0.01	0.267
CUE high gt low	-0.01	-0.02 – -0.00	0.045
stim con linear	-0.03	-0.04 – -0.01	<0.001
event02 expect angle	-0.00	-0.00 – 0.00	0.615
stim con quad	-0.01	-0.02 – 0.00	0.065
CUE high gt low * stim con linear	-0.03	-0.05 – -0.00	0.048
CUE high gt low * event02 expect angle	0.00	-0.00 – 0.00	0.107
stim con linear * event02 expect angle	0.00	-0.00 – 0.00	0.144
CUE high gt low * stim con quad	-0.02	-0.04 – 0.00	0.107
event02 expect angle * stim con quad	0.00	-0.00 – 0.00	0.255
(CUE high gt low * stim con linear) * event02 expect angle	0.00	-0.00 – 0.00	0.173
(CUE high gt low event02 expect angle) stim con quad	0.00	-0.00 – 0.00	0.527
Random Effects
σ²	0.01
τ₀₀ _sub	0.00
τ₁₁ _{sub.CUE_high_gt_low}	0.00
ρ₀₁ _sub	1.00
N _sub	70
Observations	2657
Marginal R² / Conditional R²	0.011 / NA

kableExtra::kable_styling(
  knitr::kable(
    eta_squared(model.cog3factor, partial = TRUE), # MODIFY
    "html"), "striped", position = "left", font_size = 12)

Parameter	Eta2_partial	CI	CI_low	CI_high
CUE_high_gt_low	0.0034828	0.95	0.0000433	1
stim_con_linear	0.0057816	0.95	0.0019197	1
event02_expect_angle	0.0000993	0.95	0.0000000	1
stim_con_quad	0.0013129	0.95	0.0000000	1
CUE_high_gt_low:stim_con_linear	0.0015124	0.95	0.0000074	1
CUE_high_gt_low:event02_expect_angle	0.0010157	0.95	0.0000000	1
stim_con_linear:event02_expect_angle	0.0008253	0.95	0.0000000	1
CUE_high_gt_low:stim_con_quad	0.0010038	0.95	0.0000000	1
event02_expect_angle:stim_con_quad	0.0005007	0.95	0.0000000	1
CUE_high_gt_low:stim_con_linear:event02_expect_angle	0.0007171	0.95	0.0000000	1
CUE_high_gt_low:event02_expect_angle:stim_con_quad	0.0001548	0.95	0.0000000	1

kableExtra::kable_styling(
  knitr::kable(
    lme.dscore(model.cog3factor, data_screen, type = "lme4"), # MODIFY
    "html"), "striped", position = "left", font_size = 12)

## boundary (singular) fit: see help('isSingular')

	t	df	d
CUE_high_gt_low	-2.0101599	1156.158	-0.1182366
stim_con_linear	-3.8777185	2585.733	-0.1525157
event02_expect_angle	-0.5035855	2554.118	-0.0199289
stim_con_quad	-1.8432482	2584.370	-0.0725164
CUE_high_gt_low:stim_con_linear	-1.9794942	2586.972	-0.0778374
CUE_high_gt_low:event02_expect_angle	1.6115221	2554.318	0.0637718
stim_con_linear:event02_expect_angle	1.4616274	2586.346	0.0574809
CUE_high_gt_low:stim_con_quad	-1.6118631	2585.596	-0.0633983
event02_expect_angle:stim_con_quad	1.1383984	2586.764	0.0447658
CUE_high_gt_low:stim_con_linear:event02_expect_angle	1.3623712	2586.508	0.0535758
CUE_high_gt_low:event02_expect_angle:stim_con_quad	0.6328144	2586.438	0.0248860

## Warning: Ignoring 86 observations

28.3 Kragel2018 ~ session * cue * stimulus_intensity

Q. Is the cue effect on Cognitive signature different across sessions? (Vicarious task only)

Quick answer: Yes, the cue effect in session 1 (for medium intensity group) is significantly different; low cues lead to higher similarity between the Cog-control brain maps. Howevere, this difference becomes non significant in session 4.

Session 1: 2 cue * 3 stimulus_intensity

Session 3: 2 cue * 3 stimulus_intensity

Session 4: 2 cue * 3 stimulus_intensity

28.4 Kragel2018 ~ outcome_rating

Q. Do higher Kragel 2018 map dotproducts values indicate higher outcome ratings? (Vicarious task only)

Yes, Higher Cog-control signature values are associated with higher outcome ratings. Interstingly, lower Cog-control signatures are associateed with lower outcome ratings

28.4.1 outcome_ratings * cue

28.4.2 outcome_ratings * stimulus_intensity * cue

28.5 Kragel2018 ~ expectation_rating

Q. What is the relationship betweeen expectation ratings & Cog-control signature? (Cognitive task only)

Quick answer: Kind of flat.