Gamefmri:Individual Differences Analysis

2008/07/14 notes

Goals

• address colinearity of chemotherapy attitudinal measures
• investigate self-report measures of fun and performance during fMRI task
• incorporate above analysis into an individual differences regression

2008/07/16 notes

Results

6 variable model

model 

Regression Coefficients = kills.damage + averTravel + CanRelDeath_pst + ChemoImpt_pst + ChemoGood_pst + ChemoVital_pst

note 

This model is susceptible to collinearity between ChemoImpt_pst and ChemoVital_pst. Also possibble collinearity with ChemoGood_pst

5 variable model

model 

Regression Coefficients = kills.damage + averTravel + CanRelDeath_pst + ChemoImpt_pst + ChemoGood_pst

note 

This model is susceptible to collinearity between ChemoImpt_pst and ChemoGood_pst

5 variable model 2

model 

Regression Coefficient = Performance + fmri_Fun + averTravel + CanRelDeath_pst + ChemoImpt_pst

doesn't seem good based on hippocampal analysis

4 variable model

model 

kills.damage + averTravel + CanRelDeath_pst + ChemoImpt_pst

5 variable model seems best

model 

kills.damage + averTravel + CanRelDeath_pst + ChemoImpt_pst + ChemoGood_pst

notes

correlation between ChemoImpt_pst and ChemoGood_pst, r = .27, p = .081

not a strong correlation.

ChemoVital_pst is too strongly correlated with ChemoImpt_pst

this model has the highest Adjusted R Squared value R2 adjusted = 0.3115, while adding additional variables makes it worse and removing variables also makes it worse.

Notes

lm(onset.2 + onset.3 ~ kills.damage + fMRIFun_pst + averTravel + CanRelDeath_pst + ChemoImpt_pst)

does not generate a stronger R squared than

lm(onset.2 + onset.3 ~ kills.damage + averTravel + CanRelDeath_pst + ChemoImpt_pst)

kills.damage is a better variable than Performance_pst

kills.damage does not correlate with Performance_pst, p ~ .8

Combined Chemo_score does not work well

Chemo_score = ChemoImpt_pst + ChemoVital_pst + ChemoGood_pst + ChemoPos_pst

Chemo_score = ChemoImpt_pst + ChemoVital_pst : 2 variable works slightly better than the 4 variable combination

Question 

What does the hopelab behavioral study say about these measures.