On effectively predicting autism spectrum disorder therapy using an ensemble of classifiers.

An ensemble of classifiers combines several single classifiers to deliver a final prediction or classification decision. An increasingly provoking question is whether such an ensemble can outperform the single best classifier. If so, what form of ensemble learning system (also known as multiple classifier learning systems) yields the most significant benefits in the size or diversity of the ensemble? In this paper, the ability of ensemble learning to predict and identify factors that influence or contribute to autism spectrum disorder therapy (ASDT) for intervention purposes is investigated. Given that most interventions are typically short-term in nature, henceforth, developing a robotic system that will provide the best outcome and measurement of ASDT therapy has never been so critical. In this paper, the performance of five single classifiers against several multiple classifier learning systems in exploring and predicting ASDT is investigated using a dataset of behavioural data and robot-enhanced therapy against standard human treatment based on 3000 sessions and 300 h, recorded from 61 autistic children. Experimental results show statistically significant differences in performance among the single classifiers for ASDT prediction with decision trees as the more accurate classifier. The results further show multiple classifier learning systems (MCLS) achieving better performance for ASDT prediction (especially those ensembles with three core classifiers). Additionally, the results show bagging and boosting ensemble learning as robust when predicting ASDT with multi-stage design as the most dominant architecture. It also appears that eye contact and social interaction are the most critical contributing factors to the ASDT problem among children.

Television and Magazine Alcohol Advertising: Exposure and Trends by Sex and Age.

OBJECTIVE: The purpose of this study was to document exposure to alcohol advertising by sex, age, and the level and type of alcohol people consume. METHOD: We use unique marketing survey data that link the media individuals consume and advertising appearing in those media. Our sample of 306,451 men and women represents the population age 18 and older living in the 48 contiguous United States between 1996 and 2009. We measure advertising exposure not with the standard expenditure data but with counts of actual advertisements people likely saw. We relate advertising exposure across groups defined by age, gender, and the amount of beer, wine, and spirits consumed. RESULTS: We found that drinkers, particularly young male drinkers, see much more alcohol advertising. Men, especially younger men, see more advertisements for alcohol of all types than do women. Their higher exposure is largely explained by sex differences in the propensity to read sports and adult magazines and to watch sports and gambling television programs. CONCLUSIONS: The evidence highlights the need to recognize, and when possible, control for the fact that a selected group of individuals is more likely to see alcohol advertising. Firms successfully place advertising on programs and in magazines viewed by youth and drinkers. To estimate whether seeing advertising causes people to drink (more), researchers need to develop clever identification strategies.

How do Medicare Advantage beneficiary payments vary with tenure?

OBJECTIVES: To compare how premiums and expected out-of-pocket medical costs (OOPC) vary with the length of time Medicare Advantage (MA) beneficiaries have been enrolled in their plans. STUDY DESIGN: Descriptive and fixed effects regression analyses. METHODS: Using linked administrative enrollment and plan data, we compared the costs of the MA plans that beneficiaries chose with the costs of other plans available to them. We show predicted values adjusted for age, gender, race/ethnicity, disability, individual health risk, presence of mental health diagnoses, health plan quality, relative size of the plan's provider network, and the number of years continuously enrolled in the same plan. To further address the possibility of bias, we included county-level fixed effects and compared results to a beneficiary-level fixed effects model. RESULTS: We found average spending on premiums and OOPC in enrolled plans exceeded such costs in the lowest-cost plan by $697 in 2013. Beneficiaries who remained in their plans for 6 or more years were most at risk of spending these higher amounts, paying $786 more than they would have spent in the lowest-cost plan compared with $552 for beneficiaries in their first year of enrollment. For each year a beneficiary remained in their same plan, their additional spending in excess of the minimum cost choice increased by roughly $50. CONCLUSIONS: MA beneficiaries could reduce their exposure to healthcare spending by switching to plans with lower premiums, although there may well be rational reasons for paying costs in excess of those of the lowest-cost plan.

This Ad is for You: Targeting and the Effect of Alcohol Advertising on Youth Drinking.

Endogenous targeting of alcohol advertisements presents a challenge for empirically identifying a causal effect of advertising on drinking. Drinkers prefer a particular media; firms recognize this and target alcohol advertising at these media. This paper overcomes this challenge by utilizing novel data with detailed individual measures of media viewing and alcohol consumption and three separate empirical techniques, which represent significant improvements over previous methods. First, controls for the average audience characteristics of the media an individual views account for attributes of magazines and television programs alcohol firms may consider when deciding where to target advertising. A second specification directly controls for each television program and magazine a person views. The third method exploits variation in advertising exposure due to a 2003 change in an industry-wide rule that governs where firms may advertise. Although the unconditional correlation between advertising and drinking by youth (ages 18-24) is strong, models that include simple controls for targeting imply, at most, a modest advertising effect. Although the coefficients are estimated less precisely, estimates with models including more rigorous controls for targeting indicate no significant effect of advertising on youth drinking.

Plasma fibrinogen is an accurate marker of disease activity in patients with polymyalgia rheumatica.

OBJECTIVE: The overall aim of this study was to establish whether plasma fibrinogen was a superior biomarker of disease activity in active PMR than the standard biomarkers, ESR and CRP. METHODS: Sixty patients with PMR were divided into active (n = 25) or inactive (n = 35) disease groups based on symptoms, physician assessment and biomarkers ESR and CRP. Plasma fibrinogen was assayed. Groups underwent assessment at baseline and 6 weeks. Disease activity as per the PMR activity score (PMR-AS) was recorded at all visits. Receiver operator curves (ROCs), predictive values and likelihood ratios were calculated for all biomarkers. RESULTS: Disease activity measures improved significantly in the active group between weeks 1 and 6 (P < 0.001). There was no significant difference between the activity scores at week 6 in the active group and the inactive group. Mean fibrinogen decreased from 5.2 to 3.5 g/l (normal <4 g/l) between weeks 1 and 6 in the active group. Mean ESR and CRP decreased from 59.6 to 24.3 mm/h (normal <30 mm/h) and 45.9 to 12.66 mg/l (normal <5 mg/l), respectively. Receiver operator curve analysis revealed fibrinogen to be more specific than either ESR or CRP for the detection of response to treatment in active PMR, with an overall sensitivity and specificity of 92% and 96%, respectively. Values above the upper limit of normal for fibrinogen, CRP and ESR were associated with likelihood ratios for active disease of 20.53, 2.9 and 2.8, respectively (P < 0.001). CONCLUSION: Plasma fibrinogen is at least as useful as CRP and ESR for the diagnosis of active PMR and more specific for confirmation of response to treatment than either ESR or CRP.

Smoking initiation and the iron law of demand.

We show, with three longitudinal datasets, that cigarette taxes and prices affect smoking initiation decisions. Evidence from longitudinal studies is mixed but generally find that initiation does not vary with price or tax. We show that the lack of statistical significance partly results because of limited policy variation in the time periods studied, truncated behavioral windows, or mis-assignment of price and tax rates in retrospective data (which occurs when one has no information about respondents' prior state or region of residence). Our findings highlight issues relevant to initiation behavior generally, particularly those for which individuals' responses to policy changes may be noisy or small in magnitude.