Limits...
Two different approaches to the affective profiles model: median splits (variable-oriented) and cluster analysis (person-oriented).

Garcia D, MacDonald S, Archer T - PeerJ (2015)

Bottom Line: Method.Results.More importantly, regardless of the approach, our findings suggest that the model mirrors a complex and dynamic adaptive system.

View Article: PubMed Central - HTML - PubMed

Affiliation: Blekinge Center of Competence, Blekinge County Council , Karlskrona , Sweden ; Department of Psychology, University of Gothenburg , Gothenburg , Sweden ; Network for Empowerment and Well-Being, University of Gothenburg , Gothenburg , Sweden ; Centre for Ethics, Law and Mental Health (CELAM), University of Gothenburg , Gothenburg , Sweden.

ABSTRACT
Background. The notion of the affective system as being composed of two dimensions led Archer and colleagues to the development of the affective profiles model. The model consists of four different profiles based on combinations of individuals' experience of high/low positive and negative affect: self-fulfilling, low affective, high affective, and self-destructive. During the past 10 years, an increasing number of studies have used this person-centered model as the backdrop for the investigation of between and within individual differences in ill-being and well-being. The most common approach to this profiling is by dividing individuals' scores of self-reported affect using the median of the population as reference for high/low splits. However, scores just-above and just-below the median might become high and low by arbitrariness, not by reality. Thus, it is plausible to criticize the validity of this variable-oriented approach. Our aim was to compare the median splits approach with a person-oriented approach, namely, cluster analysis. Method. The participants (N = 2, 225) were recruited through Amazons' Mechanical Turk and asked to self-report affect using the Positive Affect Negative Affect Schedule. We compared the profiles' homogeneity and Silhouette coefficients to discern differences in homogeneity and heterogeneity between approaches. We also conducted exact cell-wise analyses matching the profiles from both approaches and matching profiles and gender to investigate profiling agreement with respect to affectivity levels and affectivity and gender. All analyses were conducted using the ROPstat software. Results. The cluster approach (weighted average of cluster homogeneity coefficients = 0.62, Silhouette coefficients = 0.68) generated profiles with greater homogeneity and more distinctive from each other compared to the median splits approach (weighted average of cluster homogeneity coefficients = 0.75, Silhouette coefficients = 0.59). Most of the participants (n = 1,736, 78.0%) were allocated to the same profile (Rand Index = .83), however, 489 (21.98%) were allocated to different profiles depending on the approach. Both approaches allocated females and males similarly in three of the four profiles. Only the cluster analysis approach classified men significantly more often than chance to a self-fulfilling profile (type) and females less often than chance to this very same profile (antitype). Conclusions. Although the question whether one approach is more appropriate than the other is still without answer, the cluster method allocated individuals to profiles that are more in accordance with the conceptual basis of the model and also to expected gender differences. More importantly, regardless of the approach, our findings suggest that the model mirrors a complex and dynamic adaptive system.

No MeSH data available.


Means in positive affect (A: “Joy”) and negative affect (B: “Sadness”) for each profile derived using the median splits and cluster analysis approaches.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4631468&req=5

fig-3: Means in positive affect (A: “Joy”) and negative affect (B: “Sadness”) for each profile derived using the median splits and cluster analysis approaches.

Mentions: At a general level the distribution of the positive affect scores are approximately normal (skewness = − .18, kurtosis = − .30). The negative affect scores are heavily skewed on the right (skewness = 1.12, kurtosis = .98). This comes primarily from the fact that within the value range of negative affect (1–5) the median (1.70) is very close to the minimum (1). See Fig. 2 for the distribution of positive and negative affect and Figs. 3A and 3B for the mean in both affectivity dimensions for each of the profiles created with the median splits and cluster approaches.


Two different approaches to the affective profiles model: median splits (variable-oriented) and cluster analysis (person-oriented).

Garcia D, MacDonald S, Archer T - PeerJ (2015)

Means in positive affect (A: “Joy”) and negative affect (B: “Sadness”) for each profile derived using the median splits and cluster analysis approaches.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4631468&req=5

fig-3: Means in positive affect (A: “Joy”) and negative affect (B: “Sadness”) for each profile derived using the median splits and cluster analysis approaches.
Mentions: At a general level the distribution of the positive affect scores are approximately normal (skewness = − .18, kurtosis = − .30). The negative affect scores are heavily skewed on the right (skewness = 1.12, kurtosis = .98). This comes primarily from the fact that within the value range of negative affect (1–5) the median (1.70) is very close to the minimum (1). See Fig. 2 for the distribution of positive and negative affect and Figs. 3A and 3B for the mean in both affectivity dimensions for each of the profiles created with the median splits and cluster approaches.

Bottom Line: Method.Results.More importantly, regardless of the approach, our findings suggest that the model mirrors a complex and dynamic adaptive system.

View Article: PubMed Central - HTML - PubMed

Affiliation: Blekinge Center of Competence, Blekinge County Council , Karlskrona , Sweden ; Department of Psychology, University of Gothenburg , Gothenburg , Sweden ; Network for Empowerment and Well-Being, University of Gothenburg , Gothenburg , Sweden ; Centre for Ethics, Law and Mental Health (CELAM), University of Gothenburg , Gothenburg , Sweden.

ABSTRACT
Background. The notion of the affective system as being composed of two dimensions led Archer and colleagues to the development of the affective profiles model. The model consists of four different profiles based on combinations of individuals' experience of high/low positive and negative affect: self-fulfilling, low affective, high affective, and self-destructive. During the past 10 years, an increasing number of studies have used this person-centered model as the backdrop for the investigation of between and within individual differences in ill-being and well-being. The most common approach to this profiling is by dividing individuals' scores of self-reported affect using the median of the population as reference for high/low splits. However, scores just-above and just-below the median might become high and low by arbitrariness, not by reality. Thus, it is plausible to criticize the validity of this variable-oriented approach. Our aim was to compare the median splits approach with a person-oriented approach, namely, cluster analysis. Method. The participants (N = 2, 225) were recruited through Amazons' Mechanical Turk and asked to self-report affect using the Positive Affect Negative Affect Schedule. We compared the profiles' homogeneity and Silhouette coefficients to discern differences in homogeneity and heterogeneity between approaches. We also conducted exact cell-wise analyses matching the profiles from both approaches and matching profiles and gender to investigate profiling agreement with respect to affectivity levels and affectivity and gender. All analyses were conducted using the ROPstat software. Results. The cluster approach (weighted average of cluster homogeneity coefficients = 0.62, Silhouette coefficients = 0.68) generated profiles with greater homogeneity and more distinctive from each other compared to the median splits approach (weighted average of cluster homogeneity coefficients = 0.75, Silhouette coefficients = 0.59). Most of the participants (n = 1,736, 78.0%) were allocated to the same profile (Rand Index = .83), however, 489 (21.98%) were allocated to different profiles depending on the approach. Both approaches allocated females and males similarly in three of the four profiles. Only the cluster analysis approach classified men significantly more often than chance to a self-fulfilling profile (type) and females less often than chance to this very same profile (antitype). Conclusions. Although the question whether one approach is more appropriate than the other is still without answer, the cluster method allocated individuals to profiles that are more in accordance with the conceptual basis of the model and also to expected gender differences. More importantly, regardless of the approach, our findings suggest that the model mirrors a complex and dynamic adaptive system.

No MeSH data available.