Solving multiple linear regression problem using artificial neural network

<span>Multiple linear regressions are an important tool used to find the relationship between a set of variables used in various scientific experiments. In this article we are going to introduce a simple method of solving a multiple rectilinear regressions (MLR) problem that uses an artificial neural network to find the accurate and expected output from MLR problem. Different artificial neural network (ANN) types with different architecture will be tested, the error between the target outputs and the calculated ANN outputs will be investigated. A recommendation of using a certain type of ANN based on the experimental results will be raised.</span>

Physical activity, body mass index, and blood progranulin in older adults: cross-sectional associations in the MAPT study

Physical activity (PA) has been shown to moderate the negative effects of obesity on pro-inflammatory cytokines but its relationship with the adipokine progranulin (PGRN) remains poorly investigated. This study aimed to examine the cross-sectional main and interactive associations of body mass index (BMI) and PA level with circulating PGRN in older adults. Five-hundred and twelve subjects aged 70 years and over involved in the Multidomain Alzheimer Prevention Trial (MAPT) Study who underwent plasma PGRN measurements (ng/ml) were included. Self –reported PA levels were assessed using questionnaires. People were classified into three BMI categories: normal weight, overweight or obesity. Further categorization using PA tertiles was used to define highly active, moderately active and low active individuals. Multiple linear regressions were performed in order to test the associations of BMI, PA level, and their interaction with PGRN levels. Multiple linear regressions adjusted by age, sex, diabetes mellitus status, total cholesterol, creatinine level and MAPT group demonstrated significant interactive associations of BMI status and continuous PA such that in people without obesity, higher PA levels were associated with lower PGRN concentrations, while an opposite pattern was found in individuals with obesity. In addition, continuous BMI was positively associated with circulating PGRN in highly active individuals but not in their less active peers. This cross-sectional study demonstrated reverse patterns in older adults with obesity compared to those without obesity regarding the relationships between PA and PGRN levels. Longitudinal and experimental investigations are required to understand the mechanisms that underlie the present findings.

Association of the Posterior Acromion Extension with Glenoid Retroversion: A CT Study in Normal and Osteoarthritic Shoulders

Posterior eccentric glenoid wear is associated with higher complication rates after shoulder arthroplasty. The recently reported association between the acromion shape and glenoid retroversion in both normal and osteoarthritic shoulders remains controversial. The three-dimensional coordinates of the angulus acromialis (AA) and acromioclavicular joint were examined in the scapular coordinate system. Four acromion angles were defined from these two acromion landmarks: the acromion posterior angle (APA), acromion tilt angle (ATA), acromion length angle (ALA), and acromion axial tilt angle (AXA). Shoulder computed tomography scans of 112 normal scapulae and 125 patients with primary glenohumeral osteoarthritis were analyzed with simple and stepwise multiple linear regressions between all morphological acromion parameters and glenoid retroversion. In normal scapulae, the glenoid retroversion angle was most strongly correlated with the posterior extension of the AA (R2 = 0.48, p < 0.0001), which can be conveniently characterized by the APA. Combining the APA with the ALA and ATA helped slightly improve the correlation (R2 = 0.55, p < 0.0001), but adding the AXA did not. In osteoarthritic scapulae, a critical APA > 15 degrees was found to best identify glenoids with a critical retroversion angle > 8 degrees. The APA is more strongly associated with the glenoid retroversion angle in normal than primary osteoarthritic scapulae.

18F-FDG and 68Ga-FAPI PET/CT for Evaluating Periprosthetic Joint Infection and Aseptic Loosening in Rabbit Models

Purpose We built a loosening model based on the original infection model of rabbit and evaluated the performance characteristics of 18F-FDG and 68Ga-FAPI in infection and loosening. Methods After surgery, the rabbits were divided into four groups, six in the control group and 10 in the loosening, S. aureus and S. epidermis groups. PET/CT and serological examination were performed every two weeks for three times. After sacrificed, micro-CT, tissue culture, pullout test and scanning electron microscope were performed. Results As for 18F-FDG, performances of control and loosening groups were similar. SUVmax of S. aureus had been consistently in the high range than that of S. epidermis. As for 68Ga-FAPI, control group had the lowest SUVmax in the second week and increased gradually. SUVmax of loosening group began exceed control group since the second week. SUVmax of S. aureus in the second week was the lowest among four group and raised as the number of weeks increased and equalled to SUVmax of S. epidermis in the sixth week. Linear regressions between SUVmax and serology showed that 18F-FDG was positively correlated with CRP and IL-6 while 68Ga-FAPI revealed negative and positive correlation with CRP and IL-6 in the second and sixth week. Besides, both SUVmax and MTV of 18F-FDG or 68Ga-FAPI were negatively correlated with BV/TV and BS/TV. Conclusion In this longitudinal observation, 68Ga-FAPI showed greater sensitivity than 18F-FDG in detecting diseases, and 68Ga-FAPI had not intestinal and muscular uptake. MTV of 68Ga-FAPI were larger than 18F-FDG, which meant that 68Ga-FAPI had the potential to define the scope of lesions more accurately. Finally, SUVmax could not differentiate loosening and infection in 68Ga-FAPI, further study about diagnostic criteria was warranted.

Sex differences in the severity and natural recovery of child PTSD symptoms: a longitudinal analysis of children exposed to acute trauma

Background Higher levels of PTSD symptoms are present among trauma-exposed females v. males in adulthood; however, much less is known about the emergence of this sex difference during development. Methods In a multi-study sample of 7–18-year-olds (n = 3397), we examined the effect of sex and age on the severity of PTSD symptoms after a single incident trauma at 1 month (T1), and on symptom change after a natural recovery period of 3 (T2) and 6 months (T3). PTSD scores were harmonised across measurement types, and linear regressions were used to determine sex and age effects, adjusting for study level variance and trauma type. Results A sex × age interaction was observed at T1 (p < 0.001) demonstrating that older age was associated with greater PTSD symptom severity in females (β = 0.008, p = 0.047), but less severe symptoms in males (β = −0.011, p = 0.014). The same pattern was observed at T2 and T3, with sex differences beginning to emerge by age 12 years. PTSD symptoms decreased naturally by ~25% at T2 with little further improvement by T3. Further, females showed a greater reduction in symptoms at T3 than males, although the same effect was not observed at T2. Conclusions Sex differences in PTSD symptoms become apparent during adolescence, due to opposing changes in susceptibility occurring in females and males with age. Understanding the factors contributing to these findings is likely to provide wider insight into sex-specific psychological vulnerability to trauma-related psychopathology.

Testing times: the association of intolerance of uncertainty and metacognitive beliefs to test anxiety in college students

Background Test anxiety has a detrimental effect on test performance but current interventions for test anxiety have limited efficacy. Therefore, examination of newer psychological models of test anxiety is now required. Two transdiagnostic psychological models of emotional disorders that can account for anxiety are the intolerance of uncertainty model (IUM) and the Self-Regulatory Executive Function (S-REF) model. Intolerance of uncertainty, the stable disposition to find uncertainty distressing, is central to the IUM, while beliefs about thinking, metacognition, are central to the S-REF model. We tested for the first time the role of both intolerance of uncertainty and metacognitive beliefs in test anxiety. Methods A cross-sectional design was used, with college students (n = 675) completing questionnaires assessing their test anxiety, intolerance of uncertainty, and metacognitive beliefs. Hierarchical linear regressions examined if intolerance of uncertainty and metacognitive beliefs were associated with test anxiety, after controlling for age and gender. Results Females reported significantly more test anxiety than males. Partial correlations, controlling for gender, found intolerance of uncertainty and metacognitive beliefs were significantly and positively correlated with test anxiety. Hierarchical linear regressions found metacognitive beliefs explained an additional 13% of variance in test anxiety, after controlling for intolerance of uncertainty. When the order of entry was reversed, intolerance of uncertainty was only able to explain an additional 2% of variance, after controlling for metacognitive beliefs. In the final regression model, gender, intolerance of uncertainty and the metacognitive belief domains of ‘negative beliefs about the uncontrollability and danger of worry’ and ‘cognitive confidence’ were all significantly associated test anxiety, with ‘negative beliefs about the uncontrollability and danger of worry’ having the largest association. Conclusions Both intolerance of uncertainty and metacognitive beliefs are linked to test anxiety, but results suggest metacognitive beliefs have more explanatory utility, providing greater support for the S-REF model. Modification of intolerance of uncertainty and metacognitive beliefs could alleviate test anxiety and help students fulfil their academic potential.

Validating A Smartphone-Based Biomarker Tool Lyfas in Screening Anxiety Disorders

Lyfas is a smartphone-based biomedical application that captures the cardiovascular autonomic biomarkers (CVb), surrogating for mental health attributes. SD1/SD2 biomarker assesses the sympathovagal balance and is considered to be a potential indicator of Lyfas anxiety score (LAS). A total of 1837 males and 973 females took Lyfas (hypersensitivity-checked) and Hamilton Anxiety Rating Scale (HAM-A) self-scoring tests. LAS has been statistically validated by Linear regressions, one-way ANOVA, t-stat, correlations (r), and Bland Altman agreement assessment with respect to HAM-A. Sensitivity, specificity, precision, accuracy, Fscore, and Youden’s index (j-stat) are computed. Results show that (i) Lyfas is not a very hypersensitive instrument (mean-variance is 0.6). (ii) It can predict HAM-A with 94.7% accuracy (R2) and is a statistically significant model (p <0.05). (iii) LAS and HAM-A are positively correlated by 97%, the t-stat value of 5.38 for the population indicates that the two instruments have a significant mean difference. (iv) Bland Altman test showcases the overall agreement of 12.95% due to different modes and scales of measurements. (v) on average, LAS is 87.78% accurate, 86.82% precise, and its’ 65.2% j-stat value proves that Lyfas is a novel industry-standard smartphone biomarker application that can be used to accurately screen anxiety disorders.

Optimised Reduction of Surgical Gloves Pinholes using Forward Search Method

This research investigates the factors that affect the existence of pinholes in surgical gloves during the manufacturing process. Since eight factors affect the existence of pinholes in surgical gloves, a two-level fractional factorial design 28-4 was used to study the main effects and the first-order interactions of the multiple variables. Multiple linear regressions are used to model the data. This paper also examines the presence of influential points in the data using the influential measures in linear regression such as Cook’s Distance, DFFITS, DFBETAS, Studentized Residual, Standardized Residual, Hadi's measure, and the robust forward search. The impact of influential points is further assessed through deletion of potential influential points and model selection using adjusted R2, information criterion, and stepwise selection to see whether these influential points significantly improved the existing model.

Too tied to fail: a multidimensional approach to social capital in crowdfunding campaigns. Evidences from Italian agri-food businesses

PurposeThe authors investigate the contributions of social capital (SC) dimensions (bridging, bonding and linking) in crowdfunding campaigns by comparing the dynamics of agri-food businesses with those of two other sectors – cultural and technological.Design/methodology/approachThe authors develop linear regressions on a proprietary data set of 5,290 projects launched on the Italian platform “Produzionidalbasso.com”, from 2014 to 2020.FindingsThe authors’ findings suggest that combining the three social capital dimensions (bridging, bonding and linking) has a more substantial overall effect on the number of backers involved in agri-food projects than in cultural and technological projects. Agri-food entrepreneurs effectively mobilize all resources embedded in the SC dimensions and therefore create the conditions to develop new ties that financially support the project.Practical implicationsAgri-food entrepreneurs may benefit from those results improving their funding strategies. Therefore, agri-food entrepreneurs can explore and exploit the instruments available on the CFD platform – video and rewards associated with the campaign – gaining more benefit from the backers involved compared with other project categories.Originality/valueThe study proposes a broader perspective regarding SC that encompasses the proponent, the company and the campaign with three different types of ties: bonding, bridging and linking. These SC dimensions can differently shape diverse sectors and this eclectic configuration can differentiate the effects of SC in crowdfunding campaigns. This study pinpoints how crowdfunding determinants change, based on project categories.