SRMP: answers for assignment 6
Part 1/2 (JASP)
Open Assignment6_data. These come from a large study (Smith et al., 2017) which tested whether ‘moral foundations’ (Haidt & Joseph, 2004) can be linked to political ideology. Australian subjects filled in a 20 item questionnaire on their ‘moral foundations’ (with four items per dimension, e.g. ‘Compassion for those who are suffering is the most crucial virtue’; six-point Likert scale, 0 = ‘strongly disagree’, 5 = ‘strongly agree’). ‘Moral foundation’ variables start with the letters MF and at the end of their labels they have a letter denoting to which dimension they belong (c = Care/Harm, f = Fairness/Cheating, l = Loyalty/Betrayal, a = Authority/Subversion, d = Disgust/Purity).
Consider the five aggregated ‘moral foundations’ variables: Care, Fairness, Loyalty, Authority, and Purity. These variables are also already summed in their according dimension, such that the higher the score, the more a participant found the moral dimension important/relevant. Imagine that you hypothesize that scores on these variables are predicted by political ideology (IDEO ranging from 1 - very liberal - to 3 - very conservative), age (AGE) and sex (SEX).
- Fit models to test for the effects of
IDEO,AgeandSexon each of the aggregated ‘moral foundations’ variables and interpret the findings. Note that you do not have to use APA format, but you should give information about the direction, significance and meaning of all effects tested. (9 points)IDEOsignificantly predicts all variables, with negative effects onCareandFairness, and positive effects on the rest. In other words, a more liberal ideology predicts a higher importance ofCareandFairness, while a more conservative ideology predicts a higher importance ofAuthority,LoyaltyandPurity. (3)AGEsignificantly and positively predicts all variables. Older participants tend to value the moral dimensions more than younger participants. (3)SEXsignificantly predicts all variables exceptAuthority: females valueCare,FairnessandPuritymore than males, while males valueLoyaltymore than females. (3)
Part 2/2 (Conceptual)
- Consider all of the assumptions of the modelling technique you used in the previous section. For one of the independent variables, checking for one of these assumptions would be redundant. Explain why it is not important to check this assumption with respect to the variable in question. (3 points)
- It is not necessary to check the assumption of linearity (1) for the
SEX(1) variable as it is only represented by two categories in the dataset. This means that the relationship betweenSEXand the moral foundations variables cannot be modelled by anything more complex than a straight line/linear model (1).
- It is not necessary to check the assumption of linearity (1) for the
- A colleague suggests you think carefully about the inclusion of
Agein your models. Why might they be concerned about this? (2 points)AgeandIDEOmight be correlated (1) which could introduce collinearity (1) into the model
- Yet another colleague expresses concern about the type-I error rate of your approach. What might underlie their concerns, and what might you do to address them? (3 points)
- Type-I errors occur when \(H_0\) is incorrectly rejected, and the chances of committing such an error increase when performing multiple hypothesis tests on related data (1) - the familywise error rate (FWE). As we run five different regression models based on data from the same participants, our chances of committing a type-I error - reporting a false positive - increase. (1)
- Applying an adjustment to the \(\alpha\) level (e.g, Bonferonni correction) we use for the five models could address the issue. (1)