Title : Correlation Matrices for " Evaluating Users ' Affect States : Towards a Study on Privacy Concerns ” Names :

Background.Nwadike et al. offered a comparison of affect measurements with PANAS-X, NOLDUS Face Reader and Microsoft Emotional Recognition under a neutral state and induced happiness and sadness in a within-subject pretest experiment. Aim. We investigate the correlation matrices of the measurement devices for happiness and sadness measurements. Method. A, the measurement scores are compared with Pearson correlation matrices and corresponding visualizations, once irrespective of conditions and once within conditions "neutral," "happy," and "sad." Results. We offer a more fine-grained analysis of the measurement devices beyond the effect sizes presented in Nwadike et al. Conclusion. The analysis shows that partially the NOLDUS Face Reader have been inconsistent with the PANAS-X manipulation check, especially in the sadness condition, the Face Reader sadness score was positively correlated with PANAS-X joviality. At the same time, the pretest presented by Nwadike et al. had a low sample size and only vouches for the need of further investigation. About the authors: Current Research: Cyber Security, Privacy and Evidence-based Methods for Security I'm a Tenured Reader in System Security (Associate Professor) at the Newcastle University. I’m the Director of the Centre for Cybercrime and Computer Security (CCCS), a UK Academic Centre of Excellence in Cyber Security Research (ACE-CSR). I'm a member of the Secure and Resilient Systems group and the Centre for Software Reliability (CSR). Suggested keywords: Privacy concerns, affect states, Face Reader, Emotional Recognition, PANAS-X Correlation Matrices for “Evaluating Users’ Affect States: Towards a Study on Privacy Concerns” Thomas Groß Newcastle University Structured Abstract Background. Nwadike et al. [4] offered a comparison of affect measurements with PANAS-X, NOLDUS FaceReader and Microsoft Emotional Recognition under a neutral state and induced happiness and sadness in a within-subject pretest experiment. Aim. We investigate the correlation matrices of the measurement devices for happiness and sadness measurements. Method. A the measurement scores are compared with Pearson correlation matrices and corresponding visualizations, once irrespective of conditions and once within conditions “neutral,” “happy,” and “sad.” Results. We offer a more fine-grained analysis of the measurement devices beyond the effect sizes presented in Nwadike et al. [4] Conclusion. The analysis shows that partially the NOLDUS Face Reader have been inconsistent with the PANAS-X manipulation check, especially in the sadness condition, the FaceReader sadness score was positively correlated with PANAS-X joviality. At the same time, the pretest presented by Nwadike et al. [4] had a low sample size and only vouches for the need of further investigation. 1 Variables We analyze the correlation matrices of the experiment conducted by Nwadike et al. [4]. The experiment has been conducted as withinsubjects study, in which participants have been in a neutral state as well as exposed to standardized stimulus videos [3] for the affects happiness and sadness. The independent variable is the condition, with three levels “neutral,” “happiness,” and “sadness.” There were six dependent variables measured drawn from three measurement devices, Positive and Negative Affect Schedule (PANAS-X) [5], NOLDUS FaceReader (FR) [2, 1], and Microsoft Emotional Recognition (ER). px_sad Sadness score of the PANAS-X self-report instrument. px_jov Joviality score of the PANAS-X self-report instrument. fr_sad Sadness score of the NOLDUS FaceReader at a defined time index. fr_happy Happiness score of the NOLDUS FaceReader at a defined time index. er_sad Sadness score of the Microsoft Emotional Recognition at a defined time index.