PhD candidate Psychological Processes @MS_Utrecht丨MSc AI & MSc Psychology grad @KU_Leuven; retired 📶 engineer丨 #ItGetsBetter believer丨من هرچه ام با تو زیباترم.
- Tweets 4,557
- Following 681
- Followers 404
- Likes 21,095
ALT Hunting high and low: On the estimation and appropriateness of cosine models for experience sampling. M.M. Haqiqatkhah and E. L. Hamaker Abstract Momentary affect collected via the experience sampling method (ESM) may be characterized by circadian rhythms. To model such repetitive patterns over 24-hour cycles researchers have used cosinor analysis, which characterizes three features, namely, the intercept, amplitude, and peak offset using a cosine function. Individual differences in these parameters are of interest as they may be related to personality traits and psychopathology. Here, we raise two critical concerns about applying cosinor models to ESM data. First, some studies have used estimation methods that can yield incorrect peak offset values. This problem can be easily countered by using the correct mathematical approach. Second, since ESM data collection is limited to waking hours, the cosinor model extrapolates its 24-hour trend over periods when no data is collected due...
ALT Distribution of the estimated peak offset parameter for items (happy, sad, PA, and NA). It compares calculations using the conventional arctangent function and the two-argument arctangent function across various starting times. Red bars represent conventional calculations, while blue bars represent the improved two-argument method. The percentages of mislocated peaks are shown above each distribution.
ALT Shapes of daytime trends modeled using a cosinor function for four individuals (Person 36, 114, 141, and 39) for the ‘happy’ variable. The horizontal dashed line indicates the MESOR, with vertical lines showing modeled troughs and peaks. Gray areas represent unmeasured windows due to sleep, while light shaded lines show extrapolated trends beyond the data.
ALT Distribution of estimated peak offset parameter stratified by daily time segments: night (00:00–06:00), wake-up (06:00–08:00), daytime (08:00–22:00), and bedtime (22:00–24:00). Bars indicate the proportion of individuals with peaks occurring in each segment for variables (happy, sad, PA, and NA). Percentages are noted for each segment.
ALT Black cat staring at an oven waiting for the food to be baked
ALT Optimizing My Sleep Around Claude Usage Limits For the past 4 weeks I have been managing my sleep schedule around maximizing the usage of my Claude Pro subscription. Every five hours your Claude session usage is reset. When I first began using Claude Code daily, this limit would often come at inconvenient times. I would be in flow, vibing with Claude on my B2B SaaS side project. Unfortunately, just as I was stepping into Claude's mind like I'm Raz, the dreaded usage warning would appear. Your limit will reset at 7am. But I was just getting this React component to hydrate correctly... Frustrated, I did what any cracked coder would do and ultrathought a thorough implementation plan to my wetware README. I've also been learning to sail, having recently rediscovered a love of being both near and on the water. Sailing is fun and an addicting mental challenge - it requires being in tune with the wind (micro) and weather (macro). Having been irresponsibly inspired by Hundred Rabbits and als
ALT Screenshot of the preprint of a paper (doi.org/mjjq) by M. M. Haqiqatkhah and E. L. Hamaker titled: Daily dynamics and weekly rhythms: A tutorial on seasonal ARMA models combined with day-of-week effects.
ALT Daily dynamics and weekly rhythms: A tutorial on seasonal ARMA models combined with day-of-week effects MH Manuel Haqiqatkhah and Ellen Hamaker Daily diary data of emotional experiences are typically modeled with a first-order autoregressive model to account for possible day-to-day dynamics. However, our emotional experiences are likely to be affected by the weekly rhythm of our activities, which may be reflected by: (a) day-of-week effects (DOWEs), where different days of the week are characterized by different means; and (b) week-to-week dynamics, where weekday-specific activities and experiences have a delayed effect on the emotions that we experience on the same weekday a week later. While DOWEs have been studied occasionally, week-to-week dynamics have been largely ignored in psychological research. To gain more insight in the various regularities that may exist in daily diary data, we begin with presenting a set of complementary visualization techniques that can help to detect
ALT Figure 1 Marginal distribution, sequence, weekly, power spectral density (PSD), autocorrelation function (ACF), and partial autocorrelation function (PACF) plots for the empirical data of six individuals in Wright et al. (2015). Note that the time index in the sequence plots corresponds to the days since the start of the study.
ALT Table 3 Frequency of selected models according to the AICc, grouped based on the dynamics, lag orders, and mean structures
ALT Hunting high and low: On the estimation and appropriateness of cosine models for experience sampling. M.M. Haqiqatkhah and E. L. Hamaker Abstract Momentary affect collected via the experience sampling method (ESM) may be characterized by circadian rhythms. To model such repetitive patterns over 24-hour cycles researchers have used cosinor analysis, which characterizes three features, namely, the intercept, amplitude, and peak offset using a cosine function. Individual differences in these parameters are of interest as they may be related to personality traits and psychopathology. Here, we raise two critical concerns about applying cosinor models to ESM data. First, some studies have used estimation methods that can yield incorrect peak offset values. This problem can be easily countered by using the correct mathematical approach. Second, since ESM data collection is limited to waking hours, the cosinor model extrapolates its 24-hour trend over periods when no data is collected due...
ALT Distribution of the estimated peak offset parameter for items (happy, sad, PA, and NA). It compares calculations using the conventional arctangent function and the two-argument arctangent function across various starting times. Red bars represent conventional calculations, while blue bars represent the improved two-argument method. The percentages of mislocated peaks are shown above each distribution.
ALT Shapes of daytime trends modeled using a cosinor function for four individuals (Person 36, 114, 141, and 39) for the ‘happy’ variable. The horizontal dashed line indicates the MESOR, with vertical lines showing modeled troughs and peaks. Gray areas represent unmeasured windows due to sleep, while light shaded lines show extrapolated trends beyond the data.
ALT Distribution of estimated peak offset parameter stratified by daily time segments: night (00:00–06:00), wake-up (06:00–08:00), daytime (08:00–22:00), and bedtime (22:00–24:00). Bars indicate the proportion of individuals with peaks occurring in each segment for variables (happy, sad, PA, and NA). Percentages are noted for each segment.
ALT The cosinor model imposes symmetry, requiring peaks and troughs to be exactly half a cycle apart, which can misrepresent trends in ESM data. This restricts its use for patterns like weekly rhythms (e.g., peaks on Saturdays, troughs on Mondays) and makes interpreting data with night gaps or uneven time spans challenging. For ESM data, researchers should consider higher-order cosinor models, skewed sinusoidal regression, periodic logistic regression, or restricted spline models. For daily diary data, simpler methods like dummy variables for weekdays/weekends may suffice.
ALT When studying ILD, models should account for recurrent patterns. The single-component cosinor model may initially seem appealing due to simplicity and interpretability but struggles with physiological data, psychological self-reports, and daily diaries. Given its limitations, researchers should explore more flexible and appropriate models for such contexts.