Recent Work: Computational Modeling of Affect

Part I: Exploratory and Confirmatory Results

• Datasets
  • Exploratory: N(subs) = 93, N(exams) = 200, N(obs) = 1219
  • Confirmatory: N(subs) = 156, N(exams) = 321, N(obs) = 2103

Dataset	Affect	predictor	Estimate	SE	p-value
Exploratory	Positive	outcome	-0.2799268	0.2746970	NA
Exploratory	Positive	PE	3.0213880	0.6386234	NA
Exploratory	Negative	outcome	-0.7100594	0.4486658	NA
Exploratory	Negative	PE	-1.7950201	0.8262051	NA
Confirmatory	Positive	outcome	0.1972835	0.2705651	0.5338107
Confirmatory	Positive	PE	2.9954688	0.4683352	0.0000001
Confirmatory	Negative	outcome	-0.6125286	0.3023054	0.0727062
Confirmatory	Negative	PE	-2.3275103	0.5668274	0.0007153

Part II: CIs for Model Parameters, and PE - Outcome contrasts

• Confirmatory PA model: Outcome term is not significantly different from zero
• PE - Outcome contrast is significant for PA and NA models

Dataset	Affect	Parameter	Estimate	SE	X2	P	Lower	Upper
Exploratory	Positive	Outcome	-0.2799268	0.2746970	1.0384393	0.3081847	-0.8183229	0.2584694
Exploratory	Positive	PE	3.0213880	0.6386234	22.3832635	0.0000022	1.7697092	4.2730668
Exploratory	Positive	PE - Outcome	3.3013148	0.7834249	17.7573907	0.0000251	1.7658302	4.8367993
Exploratory	Negative	Outcome	-0.7100594	0.4486658	2.5046292	0.1135122	-1.5894281	0.1693094
Exploratory	Negative	PE	-1.7950201	0.8262051	4.7202275	0.0298098	-3.4143523	-0.1756879
Exploratory	Negative	PE - Outcome	-1.0849608	1.1430269	0.9009800	0.3425191	-3.3252523	1.1553308
Confirmatory	Positive	Outcome	0.1972835	0.2705651	0.5316651	0.4659077	-0.3330143	0.7275814
Confirmatory	Positive	PE	2.9954688	0.4683352	40.9087340	0.0000000	2.0775487	3.9133889
Confirmatory	Positive	PE - Outcome	2.7981853	0.6423043	18.9789082	0.0000132	1.5392919	4.0570786
Confirmatory	Negative	Outcome	-0.6125286	0.3023054	4.1054525	0.0427452	-1.2050362	-0.0200210
Confirmatory	Negative	PE	-2.3275103	0.5668274	16.8609287	0.0000402	-3.4384716	-1.2165489
Confirmatory	Negative	PE - Outcome	-1.7149817	0.7945313	4.6590459	0.0308904	-3.2722343	-0.1577290

Part III: Comparing ‘Outcome-Only’ and ‘PE + Outcome’ Models

• inclusion of PE term significantly improves model fit (over outcome-only models)

Dataset	Affect	Model	Df	AIC	BIC	logLik	deviance	Chisq	Chi Df	Pr(>Chisq)
Exploratory	Positive	Outcome	12	9428.141	9489.411	-4702.071	9404.141	NA	NA	NA
Exploratory	Positive	Outcome + PE	22	9363.921	9476.249	-4659.961	9319.921	84.22011	10	0
Exploratory	Negative	Outcome	12	9957.264	10018.533	-4966.632	9933.264	NA	NA	NA
Exploratory	Negative	Outcome + PE	22	9909.290	10021.617	-4932.645	9865.290	67.97405	10	0
Confirmatory	Positive	Outcome	12	16598.659	16666.472	-8287.329	16574.659	NA	NA	NA
Confirmatory	Positive	Outcome + PE	22	16478.026	16602.351	-8217.013	16434.026	140.63235	10	0
Confirmatory	Negative	Outcome	12	16968.494	17036.308	-8472.247	16944.494	NA	NA	NA
Confirmatory	Negative	Outcome + PE	22	16867.960	16992.285	-8411.980	16823.960	120.53415	10	0

Part IV: Model Fits for Sample Subject

• The paneled plot below depicts observed PA and NA (green points) for an exploratory subject, plus the following model fits
• orange: Affect ~ PE + Outcome
• blue: Affect ~ Outcome