An empirical evaluation of the Random Forests classifier models for variable selection in a large-scale lung cancer case-control study

Random Forests™ is reported to be one of the most accurate classification algorithms in complex data analysis. It shows excellent performance even when most predictors are noisy and the number of variables is much larger than the number of observations. In this thesis Random Forests was applied to a large-scale lung cancer case-control study. A novel way of automatically selecting prognostic factors was proposed. Also, synthetic positive control was used to validate Random Forests method. Throughout this study we showed that Random Forests can deal with large number of weak input variables without overfitting. It can account for non-additive interactions between these input variables. Random Forests can also be used for variable selection without being adversely affected by collinearities. ^ Random Forests can deal with the large-scale data sets without rigorous data preprocessing. It has robust variable importance ranking measure. Proposed is a novel variable selection method in context of Random Forests that uses the data noise level as the cut-off value to determine the subset of the important predictors. This new approach enhanced the ability of the Random Forests algorithm to automatically identify important predictors for complex data. The cut-off value can also be adjusted based on the results of the synthetic positive control experiments. ^ When the data set had high variables to observations ratio, Random Forests complemented the established logistic regression. This study suggested that Random Forests is recommended for such high dimensionality data. One can use Random Forests to select the important variables and then use logistic regression or Random Forests itself to estimate the effect size of the predictors and to classify new observations. ^ We also found that the mean decrease of accuracy is a more reliable variable ranking measurement than mean decrease of Gini. ^

A path analysis of an after-school health and exercise program for elementary school children in El Paso, Texas

As schools are pressured to perform on academics and standardized examinations, schools are reluctant to dedicate increased time to physical activity. After-school exercise and health programs may provide an opportunity to engage in more physical activity without taking time away from coursework during the day. The current study is a secondary data analysis of data from a randomized trial of a 10-week after-school program (six schools, n = 903) that implemented an exercise component based on the CATCH physical activity component and health modules based on the culturally-tailored Bienestar health education program. Outcome variables included BMI and aerobic capacity, health knowledge and healthy food intentions as assessed through path analysis techniques. Both the baseline model (χ2 (df = 8) = 16.90, p = .031; RMSEA = .035 (90% CI of .010–.058), NNFI = 0.983 and the CFI = 0.995) and the model incorporating intervention participation proved to be a good fit to the data (χ2 (df = 10) = 11.59, p = .314. RMSEA = .013 (90% CI of .010–.039); NNFI = 0.996 and CFI = 0.999). Experimental group participation was not predictive of changes in health knowledge, intentions to eat healthy foods or changes in Body Mass Index, but it was associated with increased aerobic capacity, β = .067, p < .05. School characteristics including SES and Language proficiency proved to be significantly associated with changes in knowledge and physical indicators. Further effects of school level variables on intervention outcomes are recommended so that tailored interventions can be developed aimed at the specific characteristics of each participating school. ^

THE APPLICATION OF PATH ANALYSIS TO ASSESSMENT OF IRON NUTRITIONAL STATUS IN MAN

Path analysis has been applied to components of the iron metabolic system with the intent of suggesting an integrated procedure for better evaluating iron nutritional status at the community level. The primary variables of interest in this study were (1) iron stores, (2) total iron-binding capacity, (3) serum ferritin, (4) serum iron, (5) transferrin saturation, and (6) hemoglobin concentration. Correlation coefficients for relationships among these variables were obtained from published literature and postulated in a series of models using measures of those variables that are feasible to include in a community nutritional survey. Models were built upon known information about the metabolism of iron and were limited by what had been reported in the literature in terms of correlation coefficients or quantitative relationships. Data were pooled from various studies and correlations of the same bivariate relationships were averaged after z- transformations. Correlation matrices were then constructed by transforming the average values back into correlation coefficients. The results of path analysis in this study indicate that hemoglobin is not a good indicator of early iron deficiency. It does not account for variance in iron stores. On the other hand, 91% of the variance in iron stores is explained by serum ferritin and total iron-binding capacity. In addition, the magnitude of the path coefficient (.78) of the serum ferritin-iron stores relationship signifies that serum ferritin is the most important predictor of iron stores in the proposed model. Finally, drawing upon known relations among variables and the amount of variance explained in path models, it is suggested that the following blood measures should be made in assessing community iron deficiency: (1) serum ferritin, (2) total iron-binding capacity, (3) serum iron, (4) transferrin saturation, and (5) hemoglobin concentration. These measures (with acceptable ranges and cut-off points) could make possible the complete evaluation of all three stages of iron deficiency in those persons surveyed at the community level. ^

SOCIAL AND BIOLOGICAL DETERMINANTS OF SHORT-TERM GROWTH VELOCITY AMONG PRESCHOOLERS OF RURAL GUATEMALA: A PATH ANALYSIS APPROACH (WEIGHT-GAIN, IRON DEFICIENCY, GASTROINTESTINAL SYMPTOMS)

This dissertation develops and tests through path analysis a theoretical model to explain how socioeconomic, socioenvironmental, and biologic risk factors simultaneously influence each other to further produce short-term, depressed growth in preschoolers. Three areas of risk factors were identified: child's proximal environment, maturational stage, and biological vulnerability. The theoretical model represented both the conceptual framework and the nature and direction of the hypotheses. Original research completed in 1978-80 and in 1982 provided the background data. It was analyzed first by nested-analysis of variance, followed by path analysis. The study provided evidence of mild iron deficiency and gastrointestinal symptomatology in the etiology of depressed, short-term weight gain. Also, there was evidence suggesting that family resources for material and social survival significantly contribute to the variability of short-term, age-adjusted growth velocity. These results challenge current views of unifocal intervention, whether for prevention or control. For policy formulations, though, the mechanisms underlying any set of interlaced relationships must be decoded. Theoretical formulations here proposed should be reassessed under a more extensive research design. It is suggested that studies should be undertaken where social changes are actually in progress; otherwise, nutritional epidemiology in developing countries operates somewhere between social reality and research concepts, with little grasp of its real potential. The study stresses that there is a connection between substantive theory, empirical observation, and policy issues. ^

A Path Forward: Evidence-Based Approaches to Educational Policy and Practice

In the complex landscape of public education, participants at all levels are searching for policy and practice levers that can raise overall performance and close achievement gaps. The collection of articles in this edition of the Journal of Applied Research on Children takes a big step toward providing the tools and tactics needed for an evidence-based approach to educational policy and practice.

Optimum Soybean Planting Date

Past research by Iowa State University has shown that the optimum planting date for soybeans, assuming favorable soil conditions, is the first week in May for the northern third of Iowa. The optimum date for the southern two thirds of Iowa is the last week of April. Given that rapidly changing soybean genetics have shown improvements in both yield and disease resistance, this trial was designed to demonstrate the planting recommendation under local conditions.

Path characteristics and colony arrival-departure statistics of king penguins (Aptenodytes patagonicus) on Possession and Kerguelen Island

Orientation based on visual cues can be extremely difficult in crowded bird colonies due to the presence of many individuals. We studied king penguins (Aptenodytes patagonicus) that live in dense colonies and are constantly faced with such problems. Our aims were to describe adult penguin homing paths on land and to test whether visual cues are important for their orientation in the colony. We also tested the hypothesis that older penguins should be better able to cope with limited visual cues due to their greater experience. We collected and examined GPS paths of homing penguins. In addition, we analyzed 8 months of penguin arrivals to and departures from the colony using data from an automatic identification system. We found that birds rearing chicks did not minimize their traveling time on land and did not proceed to their young (located in creches) along straight paths. Moreover, breeding birds' arrivals and departures were affected by the time of day and luminosity levels. Our data suggest that king penguins prefer to move in and out of the colony when visual cues are available. Still, they are capable of navigating even in complete darkness, and this ability seems to develop over the years, with older breeding birds more likely to move through the colony at nighttime luminosity levels. This study is the first step in unveiling the mysteries of king penguin orientation on land.

Pollen record from Golkow, Poland

Janczyk-Kopikowa (1966): The series of the organic deposits, developed in the vicinity of Golkow near Warsaw as oil shales and peats, was laid down in a grough valley and now rests on the deposits of the Middle Polish Glaciation (Riss). The organic deposits are overlain by the fluviale deposits of the North Polish Glaciation (Würm). The locality Golkow occurs beyond the extent of the continental glacier of this glaciation. Polen analysis completed by microfloristic examinations allows to determine the age of the organic series that is thought to be Eemian. The pollen diagram from Golkow does not call in question the stratigraphical position of the deposits investigated mainly due to its characteristic features such as minimum content of coniferous trees in the climatic optimum - about 5%, high percentage of Corylus - 77.5% and well developed phase of hornbeam. It may be well compared with other Eemian diagrams from the area of Poland and reveals much similar features. The development of vegetation at Golkow has depended upon the prevailing climate. At first, the cool climate brings about the development of plants having small thermal requirements. Here belong thin, park-like forests with pine and birch (Pinus, Betula) accompanied by the heliophilic plants such as Hippohäe and Ephedra. Improvement of climate that becomes warm and humid provides for development of deciduous forests prevailing in the climatic optimum, of the interglacial. Decrease of temperature causes a repeated change in the type of forest. This latter changes into coniferous forest with prevailing spruce (Picea) and fir (Abies) at the beginning, and then with pine (Pinus) and birch (Betula). During the Eemian Interglacial, the development of plants at Golkow terminates with a new and long-lasting predominance of pine-birch forests. However, such a longevity may be apparent only. Apparent character of this phenomenon is proved by a fact that the pollen spectra of the warm climatic periods have found their reflex in the oil shale that increased considerably slower than the layers off feebly decomposed peat evidencing the existence of cool pine-birch forests from the decline of the Interglacial. The water basin, in which the polen grains were laid down from surrounding plants is characterized by a calm sedimentation as proved by the occurrence of the oil shale. An insignificant water flow left behind some thin sand laminae. The not too deep basin becomes shallower owing to the growing water vegetation, and marshy vegetation. The growing of the plants causes a complete shallowing of the basin and formation of peat bog in situ, as proved by the peat beds occurring in the section. ---- Gadomska (1966): In the vicinity of Golków a series of organic deposits occurs amounting to 6.5-9.3 m in thickness, and consisting of oil shales, lacustrine silts and sands, as well as peats and peaty silts. The organic deposits fill up an old, small, but fairly deep lake basin, probably of finger-lake origin. It may be seen to-day as a slight lowering of the relief, filled up with soaked ground, stretching from north to south. On the basis of palaeobotanical examinations the organic deposits considered are of Eemian Interglacial age (Z. Janczyk-Kopikowa, 1063). The lower part of the organic series consists of a compact oil shale horizon, the maximum thickness of which may attain up to 8 m. The oil shales contain particularly in their upper part, numerous intercalations of arenaceous silts, dark grey or black in colour, or of sands mainly of lacustrine provenance. At the top of the oil shales are found peats, up to 2.5 m in thickness, covered by black, humus silts with numerous plant remains. The Eemian Interglacial deposits are covered by a series of fluviatile sands belonging partly to the Baltic Glaciation (bottom part of the series), partly to the Holocene (top part of the series). The thickness of the sands is 0.5-3.7 m. Higher up, there are found the Holocene and present-day deposits developed as clayey alluvion, or arenaceous slide rocks, or arenaceous-silty soil.

Pollen record and age determination of a sediment profile from Frengstadsetra, Norway

Innerdalen was once a mountain valley (ca. 780 m a.s.l.) with birch forests, bogs and several summer farms. Today it is a 6.5 km**2 artifical lake. In 1980 and 1981 archaeological and palynological investigations were carried out due to the hydroelectric power plans. Radiocarbon dated pollen diagrams from 9 different localities in Innerdalen provide information on a mountain environment which has been exploited to varying degrees by human groups for thousands of years. In the Birch Zone, ca. 9500-8500 years B.P., the deglaciated surface is vegetated by the normal sequence of pioneering species, first show-bed communities, then shrub/dwarf-shrub communities, and finally a birch forest community. In the Pine Zone, ca. 8500-7500 years B.P., the mixed Birch-Pine forest which prevailed at the end of the Birch Zone is replaced by a dense pine forest. The tree limit was higher than it is today. In the Alder Zone, ca. 7500-4000 years B.P., the newly arrived alder gradually succeeded pine, particularily on good soils. This alder forest has a modem analog in the pre-alpine gray alder forests in Norway. In the last part of the Alder Zone, ca. 6000-4000 years B.P., elm and hazel are nominally present on particularily rich soils, marking the edaphic and climatic optimum in Innerdalen. During this time the first evidence of human impact on the vegetation is apparent in the pollen diagrams. At both Sætersetra in the south of the valley and Liabekken in the north, forest clearance and the development of grazed grass meadows is documented, and human impact continues until the present. The Herb Zone, ca. 4000 years B.P. to 1600 A.D., is characterized by the rapid decline of alder. The forest is increasingly open, and bog formation is initiated. The sub-alpine belt of birch forest is established, probably due to the shift to a cooler, moister climate. Human activity can also have influenced the vegetational changes, although at 4 of the localities human activity also is first apparent after the alder decline. Some localities show measurably less human impact on the vegetation ca. 2600-2000 years B.P. Grazing intensity increases ca. 2000 years B.P. At the end of the Herb Zone rye and barley pollen is registered at Sætersetra and Flonan, indicating contact between the grazing activities of Innerdal and grain cultivation activities outside the valley. The Spruce Zone, ca. 1600 A.D. to the present, does not begin synchronously since the presence of long-distance transported spruce pollen at a locality is entirely dependent on the density of the vegetation ie. degree of human impact. The youngest spruce rise is ca. 1500 A.D. at Røstvangen, when summerfarming is initiated. Summerfarming activities in Innerdal produce an increasingly open landscape. Rye and barley pollen at several localities may indicate limited local cultivation, but is more likely long-distance transport via humans and domesticated animals from cultivated areas outside Innerdalen.