Datasets:

SouthernCrossAI
/

Arxiv

Dataset card Data Studio Files Files and versions Community

Split (1)

train · 132k rows

Filename stringlengths 22 64	Paragraph stringlengths 8 5.57k
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	Missing data mechanisms are important since older methods of handling missing data (e.g. available case analysis) assume MCAR missingness while modern techniques to handle missing data such as maximum likelihood esti- mation (MLE) and multiple imputation assume only MAR.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	fatal liver cancer (El-Serag, 2012). HCV causes the inﬂammation of the liver and over time this can lead to liver diseases like cirrhosis and ﬁbrosis. HCV is a blood borne virus and in developed countries it is often transmitted Indeed, in Australia in through risky behaviour such as sharing needles. 2007 (Razali et al., 2007), 90% of the new and 80% of the existing infections are transmitted in this way.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	While there are advances in the treatment of people with HCV, the hu- man costs of HCV, in terms of reduction of quality of life and well being and through occupational and social discrimination and isolation, remain signif- icant (Vietri et al., 2013). Guidelines for the care and treatment of HCV have recently been updated (Organization et al., 2018). The ﬁnancial costs of the virus for medical and hospital care, lost productivity, and the need for social support are also an increasing proportion of healthcare expenditure in Australia. While the diagnosis of HCV is most conﬁdently made through the immunoassay (for Study of Liver et al., 2014), having a powerful predictive model will help focus on early detection of HCV infection particularly for cases where the speciﬁc immunoassay has not been ordered.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	Analysis of multivariate data is often negatively aﬀected by incompleteness. Reduced statistical power, increase in standard errors, complications in data handling and analysis, and introduction of bias are some of the common problems associated with missing data (Horton and Lipsitz, 2001).
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	dard errors reﬂect the true variation and uncertainty in the data better than the estimates obtained by deleting any observations with missing values, also known as available case analysis. MI is less biased than single imputation methods, increases the eﬃciency of estimation and also preserves the vari- ability in the dataset. Hence, it is the preferred method to ﬁll in missing values in datasets such as the one used in this study. It should be noted that the goal of imputation is not to replace or recover these missing values from the dataset, as they are unknown, but to produce valid and analytical results in the presence of the missing values.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	Despite its conception in the 1980s, MI has only recently come into gen- eral use across medical research due to advances in statistical computation. Even with the evolution of MI and its statistical advantages, the technique has had limited application in laboratory diagnostics. The use of MI in clin- ical chemistry has been supported (Janssen et al., 2009; Waljee et al., 2013) when values of a variable in an existing prediction model are missing.The study presented herein extends this previous approach by imputing variables with an unknown relationship to the outcome of interest.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	The primary goal of this study is to show the importance of multiple im- putation when dealing with missing values in routinely collected pathology data. A secondary goal of this paper is to contribute to knowledge of lab- oratory prediction of HCV by identifying novel combinations of routinely measured biomarkers that are highly predictive of HCV. This focus on an existing resource which provides a low-cost approach to knowledge discovery is important in the context of low-middle income countries, where laboratory resources may be limited (Shang et al., 2013).
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	both (Figure 1). There were 4,296 patients for whom 60% or more of the other assays (Table 1) were missing, and these were removed before analy- sis commenced due to the large proportion of missingness. Another 3,546 individuals were missing a HCV immunoassay (HepC) outcome despite a re- quest having been made and so excluded from the analysis. The result of the HCV immunoassay is recorded as positive if the test is positive to the pres- ence of HCV anti-bodies. The same dataset has been analysed previously (Richardson and Lidbury, 2013, 2017) to ascertain the interaction between virus, outcome, pre-processing and method on the performance of decision tree ensembles and support vector machines.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	For access to de-identiﬁed patient data, this study had human ethics ap- proval granted by The Australian National University Human Ethics Com- mittee (2012/349) and the ACT Health Human Research Ethics Committee (ETHLR.11.016).
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	This is a cross-sectional dataset where each individual only appears once with a non-monotone pattern of missingness. Percentage missingness is calculated for each variable and we examine the pattern of missingness visually to get an impression of the extent of incompleteness. These patterns of missingness inﬂuences the amount of information that can be transferred between vari- ables (Van Buuren, 2012). For example, imputations can be more precise if complete information is available for the other variables for the observations that are to be imputed.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	Table 1: Diagnostic pathology variables included in the regression analysis, with summary statistics (n = 10,774).
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	b=1 The value of Ia depends on the proportion of missing data in the variable Xa. This means that if Xa has no missing values, Ia = 1. The Inﬂux coeﬃcient (Ia) is the number of variable pairs (Xa, Xb), where Xa is missing and Xb is observed. Thus, a high inﬂux implies how well the variable is connected with the observed data.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	Imputation, Analysis and Pooling. Final inference on the regression coeﬃ- cient estimates is made on the pooled result using Rubin’s combination rule (Little and Rubin, 2002). Variables with an Outﬂux coeﬃcient above 0.95 i.e. variables that occurred in the top left corner of the ﬂux plot, are used as predictors in the imputation model. We employ a stepwise procedure by calculating the number of times a variable occurred a predictor in the im- putation model. This method is also referred to as the impute then select method (Yang et al., 2005). A supermodel is constructed using the variables that occurred across all imputed datasets. For all other variables, we use the multivariate Wald test and the likelihood ratio test to determine whether the variable should be included in the ﬁnal model.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	Multiple Imputation using Chained Equations (MICE) (Van Buuren, 2012) has emerged in the literature as a routine method for handling missing data in both continuous and binary variables. Descriptive statistics (mean and standard deviation) for all iterations are used to obtain a plot to check im- putation convergence. If the mean and standard deviation of the imputed variables appear to have settled at particular values, the imputation process is deemed to have converged. There is a lack of formal tests of convergence (Su et al., 2011), so the plots of statistics such as the mean and standard deviation are used to provide visual information about convergence. A back- wards step wise model selection procedure is employed in the analysis phase after imputations.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	All analysis was performed using the statistical software R (version 3.4.0) (R Core Team, 2021). The package ‘mice’ (Van Buuren and Groothuis- Oudshoorn, 2011) were used for multiple imputation. The package ‘ROCR’ (Sing et al., 2005) is used to produce Receiver Operating Characteristic (ROC) curves to test the predictive ability of the imputed models and to calculate the area under the curve (AUC). The packages ‘nortest’ (Gross and Ligges, 2015) and ‘psych’ (Revelle, 2022) are used to perform the Anderson- Darling test of normality and to obtain summary statistics respectively. The package ‘MASS’ (Venables and Ripley, 2002) is used to perform χ2 and Mann- Whitney U tests.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	Table 2: Diagnostic pathology variables included in the regression analysis, with summary statistics (n = 10,774).
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	predictors, transformations were applied to ensure predictors were of similar order of magnitude across analysis methods. Complex transformations (like Box-Cox power transformations) were not considered for the ﬁnal model for clarity of interpretation. Finally, logistic regression is used to predict HCV status on the basis of age, sex and the 23 biomarkers.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	There were over 100 diﬀerent patterns of missingness in the predictor vari- ables across the 14,320 patients (Figure 2a). There were 7,820 individuals (54.6%) with complete data on age, sex and all 23 biomarkers. Percentages of missingness (Figure 2b, Table 2) occurs in one of three ways. For Sodium, Potassium, creatinine (Crea) and Urea, there is approximately 15% missingness, whether or not a patient was HCV positive. For serum al- bumin (ALB), alanine aminotransferase (ALT), alkaline phosphatase (ALP), gamma glutamyl transferase (GGT) and total serum bilirubin (TBil), there is about 15% missingness amongst HepC negative, and less than 1% missingness amongst HepC positive. For the remainder (Age, Sex, platelets (Plt), mono- cytes (Mono), eosinophils (Eos), basophils (Bas), lymphocytes (Lymph)), there was only occasional missingness of data.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	The area under the ROC curve (AUROC) (Figure 5a) for the available cases logistic regression model on the validation dataset was 72%; namely, the model has a 72% chance of correctly classifying individuals as having a pos- itive HepC assay. The model identiﬁed the variables Age, Sex, Sodium, RDW, Potassium, log(ALT), log(ALP), log(Crea), log(TBil), log(Urea) and sqrt(Lymph) as signiﬁcant (p < 0.05; see Table 3).
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	comprise: Age, Sex, Hb, RCC, MCV, Hct, RDW, Mch, MCHC, Plt ,WCC, Mono, Eos, Bas, Neut and Lymph.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	We ﬁt a stepwise logistic model to predict HepC and count the number of times each variable occurred in the imputation model in each of the imputed datasets. This was done for ﬁve, and 20 imputations. We observe that vari- ables that did not occur in any of the imputation models when there were ﬁve imputations, occur in lower frequencies when there were 20 imputations.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	Table 4: Tabulation of the number of occurrences in imputation model across all imputed datasets.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	Variables that occurred in all imputation models were considered in the supermodel.For other variables, the likelihood ratio test was applied to iden- tify if the variable should be included in the ﬁnal model. Variables with p-value < 0.05 were selected in the ﬁnal supermodel. Table 4 shows the results of the tabulation for number of occurrences of each variable when number of imputations were equal to 5 and 20.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	The parameter estimate from the multiple imputation methods are au- tomatically pooled and presented in Table 5. The AUROCs of the ﬁve-fold imputed models have a mean of 72% (SD 0.003). We increased the num- ber of imputations from 5 to 20 and observed variable selection was more consistent, reducing the randomness in the variable selection process. The AUROCs of the twenty-fold imputed models have a mean of 71% (SD 0.002).
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	Discussion of the results will involve both the meaning of the results for mul- tiple imputation in general, and the results for the diagnosis of hepatitis C in particular. The focus of the study is to improve the diagnostic process by developing a powerful predictive model for HepC without discarding any information collected during laboratory tests due to the presence of missing values, and using the completed dataset for developing this model.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	The percentage of missing values displayed three patterns (see Results). Biomarkers with around 15% missingness included sodium, potassium, cre- atinine and urea which are associated with the Urea-Electrolytes-Creatinine battery of tests. These are routinely requested for most laboratory investi- gations. Biomarkers such as ALB, ALT, ALP, GGT and TBil and displayed around 15% missingness for those found HCV negative, and less than 1% missingness for those found HCV positive. These are routine liver function tests and would be primary markers sought when HepC is suspected. The remaining biomarkers are from the full blood count and are routinely re- quested, and so records are complete for almost all individuals. A recent paper (Lidbury, 2018) has been published on this strategy, but with the goal of avoiding painful and expensive liver biopsy.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	a HCV infection. In addition to early detection, the advantage of using qual- ity routine pathology data as a predictive model, where immunoassay is not easily available is a key outcome especially when handling population health issues for rural or remote areas in third world countries.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	The regression models suggest that increased levels of ALT, RDW and lym- phocytes increased the odds of detecting HCV infection. All models pre- sented include the variable ALT which implies that ALT is one of the most useful routine test predictors of HCV infection. The ﬁnding for ALT coincides with current medical practice (Holmes et al., 2013), as cases with elevated ALT levels are more likely to be infected with HCV. However, ALT elevation is not speciﬁc for viral hepatitis [(Kwo et al., 2017) and so an increased level of ALT in the blood is used as a guide to request speciﬁc second-tier tests that may include HepC immunoassay.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	RDW has also been previously identiﬁed as having a relationship with HCV in a small-scale Chinese in-patient study (Shang et al., 2013); the current analysis supports that ﬁnding in a much larger Australian community cohort. Lymphocytes have also been identiﬁed as playing a role in HCV infection in a systematic review of epidemiological studies (He et al., 2016).
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	Age is another strong predictor of HCV status, with increasing age decreas- ing the chance of being HCV positive. Evidence from Europe (Cacoub et al., 2016) and Australia (Faustini et al., 2010) supports this. The increased prevalence in this study of HCV diagnosis in younger patients suggests that clinicians could use this evidence to be proactive in seeking information on the key variables identiﬁed in this study in the younger age groups. Although the laboratory supplying the data is housed on a hospital campus, the organ- isation provides pathology services to the public as well as inpatients, hence the sample contains many community-living persons with a wide range of clinical indications leading to the request for an HCV immunoassay.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	of this study is that clinicians are likely to use their clinical judgement to order some tests and not others. This clinical judgment is now also being enhanced by data mining (Lidbury et al., 2015). Non-clinical characteris- tics can also drive choice of tests requested (Smellie et al., 2002). Practice characteristics and other clinical notes are not captured in the available data.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	A second and related limitation of this multiple imputation analysis is that were MAR to be assumed, the parameter estimates from the imputed data are very sensitive to the model employed for the probability of response (Lit- tle and Rubin, 2002). The probability of response can be modelled by using linear regression models for continuous variables, logistic regression models for binary variables, and multinomial logistic regression models for categori- cal variables with more than two classes.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	A ﬁnal limitation of this analysis concerns the rarity of the outcome: only 3% of subjects tested positive for HCV. Whilst the rarity of the outcome supports our view that inclusion bias is likely to be small, it does raise issues around the stability of classical estimation algorithms. Logistic regression for rare events may require the use of penalised likelihood instead of maximum likelihood (King and Zeng, 2001).
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	oratory data has identiﬁed combinations of routinely measured biomarkers that are integral to the prediction of HCV infection.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	Acknowledgements: The authors wish to thank Dr Gus Koerbin and staﬀ at ACT Pathology for their support of this project. They also thank Mr Sheikh Faisal who undertook the initial analysis of this data as part of his Master’s degree research.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	Financial support: This research did not receive any speciﬁc grant from funding agencies in the public, commercial, or not-for-proﬁt sectors.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	Blach, S., S. Zeuzem, M. Manns, I. Altraif, A.-S. Duberg, D. H. Muljono, I. Waked, S. M. Alavian, M.-H. Lee, F. Negro, et al. (2017). Global preva- lence and genotype distribution of hepatitis c virus infection in 2015: a modelling study. The lancet Gastroenterology & hepatology 2 (3), 161–176.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	Cacoub, P., C. Comarmond, F. Domont, L. Savey, A. C. Desbois, and D. Saadoun (2016). Extrahepatic manifestations of chronic hepatitis c virus infection. Therapeutic advances in infectious disease 3 (1), 3–14.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	Dempster, A. P., N. M. Laird, and D. B. Rubin (1977). Maximum likelihood from incomplete data via the em algorithm. Journal of the Royal Statistical Society: Series B (Methodological) 39 (1), 1–22.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	Faustini, A., P. Colais, E. Fabrizi, A. M. Bargagli, M. Davoli, D. Di Lallo, A. Di Napoli, P. Pezzotti, C. Sorge, R. Grillo, et al. (2010). Hepatic and extra-hepatic sequelae, and prevalence of viral hepatitis c infection esti- mated from routine data in at-risk groups. BMC infectious diseases 10 (1), 1–13.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	for Study of Liver, E. A. et al. (2014). Easl clinical practice guidelines: management of hepatitis c virus infection. Journal of hepatology 60 (2), 392–420.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	Gross, J. and U. Ligges (2015). Package ‘nortest’: Five omnibus tests for testing the composite hypothesis of normality. Dortmund, Germany: Dort- mund University. R package version 1.0.4.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	He, Q., Q. He, X. Qin, S. Li, T. Li, L. Xie, Y. Deng, Y. He, Y. Chen, and Z. Wei (2016). The relationship between inﬂammatory marker levels and hepatitis c virus severity. Gastroenterology Research and Practice 2016.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	Horton, N. J. and S. R. Lipsitz (2001). Multiple imputation in practice: com- parison of software packages for regression models with missing variables. The American Statistician 55 (3), 244–254.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	Janssen, K. J., Y. Vergouwe, A. R. T. Donders, F. E. Harrell Jr, Q. Chen, D. E. Grobbee, and K. G. Moons (2009). Dealing with missing predictor values when applying clinical prediction models. Clinical chemistry 55 (5), 994–1001.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	Kwo, P. Y., S. M. Cohen, and J. K. Lim (2017). Acg clinical guideline: evaluation of abnormal liver chemistries. Oﬃcial journal of the American College of Gastroenterology— ACG 112 (1), 18–35.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	Lidbury, B. A. (2018). Predicting liver disease post hepatitis virus infection: In silico pathology and pattern recognition. EBioMedicine 35, 10–11.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	Lidbury, B. A., A. M. Richardson, and T. Badrick (2015). Assessment of machine-learning techniques on large pathology data sets to address assay redundancy in routine liver function test proﬁles. Diagnosis 2 (1), 41–51.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	Organization, W. H. et al. (2018). Guidelines for the care and treatment of persons diagnosed with chronic hepatitis C virus infection. World Health Organization.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	Razali, K., H. H. Thein, J. Bell, M. Cooper-Stanbury, K. Dolan, G. Dore, J. George, J. Kaldor, M. Karvelas, J. Li, et al. (2007). Modelling the hepatitis c virus epidemic in australia. Drug and alcohol dependence 91 (2- 3), 228–235.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	Revelle, W. (2022). psych: Procedures for Psychological, Psychometric, and Personality Research. Evanston, Illinois: Northwestern University. R pack- age version 2.2.3.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	Richardson, A., B. M. Signor, B. A. Lidbury, and T. Badrick (2016). Clinical chemistry in higher dimensions: machine-learning and enhanced predic- tion from routine clinical chemistry data. Clinical biochemistry 49 (16-17), 1213–1220.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	Infection status outcome, machine learning method and virus type interact to aﬀect the optimised prediction of hepatitis virus immunoassay results from routine pathology laboratory assays in unbalanced data. BMC bioinformatics 14 (1), 1–8.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	Richardson, A. M. and B. A. Lidbury (2017). Enhancement of hepatitis virus immunoassay outcome predictions in imbalanced routine pathology data by data balancing and feature selection before the application of support vector machines. BMC medical informatics and decision making 17 (1), 1–11.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	Rubright, J. D., R. Nandakumar, and J. J. Gluttin (2014). A simulation study of missing data with multiple missing x’s. Practical Assessment, Research, and Evaluation 19 (1), 10.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	in chinese patients by pathology data mining. Journal of medical virol- ogy 85 (8), 1334–1339.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	Sing, T., O. Sander, N. Beerenwinkel, and T. Lengauer (2005). Rocr: visu- alizing classiﬁer performance in r. Bioinformatics 21 (20), 3940–3941.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	Is clinical practice variability the major reason for diﬀerences in pathology requesting patterns in general practice? Journal of clinical pathology 55 (4), 312–314.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	Stanaway, J. D., A. D. Flaxman, M. Naghavi, C. Fitzmaurice, T. Vos, I. Abubakar, L. J. Abu-Raddad, R. Assadi, N. Bhala, B. Cowie, et al. (2016). The global burden of viral hepatitis from 1990 to 2013: ﬁndings from the global burden of disease study 2013. The Lancet 388 (10049), 1081–1088.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	Su, Y.-S., A. Gelman, J. Hill, and M. Yajima (2011). Multiple imputation with diagnostics (mi) in r: Opening windows into the black box. Journal of Statistical Software 45, 1–31.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	Van Buuren, S. and K. Groothuis-Oudshoorn (2011). mice: Multivariate imputation by chained equations in r. Journal of statistical software 45, 1–67.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	Vietri, J., G. Prajapati, and A. C. El Khoury (2013). The burden of hepatitis c in europe from the patients’ perspective: a survey in 5 countries. BMC gastroenterology 13 (1), 1–8.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	Von Hippel, P. T. (2007). 4. regression with missing ys: an improved strategy for analyzing multiply imputed data. Sociological Methodology 37 (1), 83– 117.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	Waljee, A. K., A. Mukherjee, A. G. Singal, Y. Zhang, J. Warren, U. Balis, J. Marrero, J. Zhu, and P. D. Higgins (2013). Comparison of imputation methods for missing laboratory data in medicine. BMJ open 3 (8), e002847.
Processed_The_Effect_of_Multiple_Imputation_of_Routine_Patho.txt	Yang, X., T. R. Belin, and W. J. Boscardin (2005). Imputation and vari- able selection in linear regression models with missing covariates. Biomet- rics 61 (2), 498–506.
Processed_The_Formation_of_Constellation_III_in_the_Large_Ma.txt	Abstract: We present a detailed reconstruction of the star-formation history of the Con- stellation III region in the Large Magellanic Cloud, to constrain the formation mechanism of this enigmatic feature. Star formation in Constellation III seems to have taken place during two distinct epochs: there is the 8-15 Myr epoch that had previously been recog- nized, but we also see strong evidence for a separate “burst” of star formation 25–30 Myr ago. The “super-supernova” or GRB blast wave model for the formation of Constella- tion III is diﬃcult to reconcile with such an extended, two-epoch star formation history, because the shock wave should have induced star formation throughout the structure simultaneously, and any unconsumed gas would quickly be dissipated, leaving nothing from which to form a subsequent burst of activity. We propose a “truly stochastic” self-propagating star formation model, distinct from the canonical model in which star formation proceeds in a radially-directed wave from the center of Constellation III to its perimeter. As others have noted, and we now conﬁrm, the bulk age gradients demanded by such a model are simply not present in Constellation III. In our scenario, the prestel- lar gas is somehow pushed into these large-scale arc structures, without simultaneously triggering immediate and violent star formation throughout the structure. Rather, star formation proceeds in the arc according to the local physical conditions of the gas. Self- propagating star formation is certainly possible, but in a truly stochastic manner, without a directed, large scale pattern.
Processed_The_Formation_of_Constellation_III_in_the_Large_Ma.txt	Shapley’s Constellation III in the Large Magellanic Cloud (LMC) is one of the most enigmatic structures in the local universe: a coherent semicircular arc span- ning several hundred parsecs, composed of thousands of bright young stars and tens of star clusters. Its reg- ularity across such a large scale deﬁes the fractal-like distributions that young stellar populations typically follow; in fact, Constellation III may be unique in this regard. In addition, Constellation III is embedded in- side the supergiant shell LMC 4, a circular hole in the LMC’s HI disk that spans more than a kiloparsec (Kim et al. 1999), and whose rim is dotted with HII regions (Meaburn 1980).
Processed_The_Formation_of_Constellation_III_in_the_Large_Ma.txt	of star clusters which must have hosted similarly strong massive-star winds, yet it contains no other structures like Constellation III. Instead, Efremov & Elmegreen (1999) favor an updated “super-supernova” idea, sug- gesting the LMC4 cavity was blown by a gamma-ray burst formed by the coalescence of an X-ray binary (perhaps ejected from the nearby rich star cluster NGC 1978), and the swept-up material subsequently formed Constellation III. Dopita et al. (1985) presented the idea that Constellation III is the result of a stochas- tic self-propagating star formation (SSPSF) process, directed in an outward radial propagation. However, recent studies of the distribution of ages in Constella- tion III have not conﬁrmed the radial age gradient re- ported by Dopita et al. (Olsen et al. 1997; Braun et al. 1997; Dolphin & Hunter 1998).
Processed_The_Formation_of_Constellation_III_in_the_Large_Ma.txt	In this paper, we present a map of the past history of star formation in the vicinity of Constellation III, in order to constrain the various formation scenarios for this unique structure. In Section 2, we brieﬂy review the photometric data and our StarFISH analysis soft- ware. In Section 3, we present a detailed map of the star formation history (SFH) throughout the Constel- lation III region. We discuss the implications of our SFH map in Section 4, and summarize the results in Section 5.
Processed_The_Formation_of_Constellation_III_in_the_Large_Ma.txt	Figure 1: A stellar ﬂux density map of a 2.5◦×2.5◦ region in the LMC, including Constellation III. The map was derived from our MCPS photome- try: each pixel’s value is proportional to the to- tal stellar ﬂux in B, V and I (for blue, green and red, respectively). Major structures and clus- ters are labeled, including Constellation III itself (dashed outline), and the approximate position of the LMC 4 supergiant shell (large dotted circle). Note that Constellation III is actually one of a few large stellar arcs in this region; our analysis will include all of these arcs.
Processed_The_Formation_of_Constellation_III_in_the_Large_Ma.txt	Figure 2: The same ﬂux-density maps as in Fig- ure 1, with our conformal grid overplotted. We de- termined the best-ﬁt SFH solution independently for each grid cell in order to generate a map of the SFH in this region.
Processed_The_Formation_of_Constellation_III_in_the_Large_Ma.txt	try and photometry for 24 million LMC stars and more than 6 million SMC stars.
Processed_The_Formation_of_Constellation_III_in_the_Large_Ma.txt	In Figure 1, we show a stellar ﬂux density map ◦ ◦ × 2.5 derived from our MCPS photometry, for a 2.5 region including Constellation III. The Figure shows that the arc traditionally known as Constellation III is actually one of at least three large stellar arcs in this region; all of these arcs lie in the interior of the LMC 4 HI supergiant shell.
Processed_The_Formation_of_Constellation_III_in_the_Large_Ma.txt	Figure 3: The six panels on the left show (B − V ) vs. V CMDs for six of our Constellation III regions (as labeled at the top of each panel; see Figure 2). The termination of the main sequence at V =12– 13 mag indicates that the youngest stars in these regions is aged 10–15 Myr (the red isochrones over- plotted in each panel). In addition, the three pop- ulations in the top row show an isolated clump of red supergiants near V = 13.5 mag, indicative of a burst of star formation activity around 30 Myr ago (the green isochrones overplotted in the top panels). This is further illustrated by the synthetic stellar populations in the rightmost column. These two populations diﬀer only in their SFH between 10 and 100 Myr: in the top panel, the stars in this age range all have an age of 30 Myr, whereas in the bottom panel, the ages are uniformly distributed between 10 and 100 Myr.
Processed_The_Formation_of_Constellation_III_in_the_Large_Ma.txt	licity, we can generate composite model CMDs that represent any arbitrary SFH. These composite model CMDs can then be quantitatively compared to the real, observed CMD, and by minimizing the diﬀerences be- tween them, the best-ﬁtting SFH solution can be ob- tained.
Processed_The_Formation_of_Constellation_III_in_the_Large_Ma.txt	In order to distinguish between the various competing theories for the formation of Constellation III, we con- structed a spatially-resolved map of the SFH of the entire 2.5◦ × 2.5◦ region. In order to maximize our ability to detect any radial or azimuthal population gradients present in these stellar arcs, we constructed a conformal grid that follows their curvature (see Fig- ure 2), and determined an independent SFH solution for the stars in each grid cell. The synthetic CMDs employed for each grid cell used extinction distribu- tions and empirical photometric error models derived directly from the grid cell’s stellar population.
Processed_The_Formation_of_Constellation_III_in_the_Large_Ma.txt	Previous analyses of the SFH of Constellation III have largely sought to determine a single characteris- tic age for diﬀerent locations in the structure, without considering the full distribution of stellar ages that makes up the true SFH. In particular, the ages as- signed have been the youngest age present. As an il- lustration that a more complete SFH is warranted for these regions, we examine the (B − V ) CMDs of eight of our Constellation III regions in Figure 3. Each of these regions shows a prominent main sequence, and by simple isochrone ﬁtting, one can determine the age of the youngest stellar population in each region (as shown by the red curves in each panel). However, these CMDs show clear evidence for a more complex SFH. The red giant branch and red clump are obvi- ous tracers of old stellar populations, but there are also supergiants present which trace star formation that occurred several tens of millions of years ago. In particular, the CMDs in the top row of Figure 3 each show an isolated knot of red supergiants with V = 13.5 mag. An isolated knot of supergiants at a common luminosity is strong evidence for an isolated burst of star formation activity in the recent history of the region, because the luminosity of a supergiant is directly and unambiguously correlated with its age (Dohm-Palmer et al. 1997). The isochrone overplotted in green in Figure 3 indicates that the supergiants in these knots are roughly 30 Myr old. Thus, by simple visual inspection of these CMDs, we can already con- clude that some of the Constellation III regions have experienced multiple, isolated bursts of star formation. Variations like these will be recovered in our StarFISH analysis, giving us a much more complete picture of the SFH of Constellation III.
Processed_The_Formation_of_Constellation_III_in_the_Large_Ma.txt	When examining Figure 4, it is important to un- derstand that SFH maps like this do not allow us to display information on the uncertainties associated with the best-ﬁt star formation rates (SFRs) in each time step. However, StarFISH does estimate these un- certainties, and they do include covariance between ad- jacent age bins. So, for example, in the map there are many regions that have a large SFR in the 10 Myr panel, but a very low SFR in the 8 Myr panel, and vice versa. The uncertainties computed by StarFISH for these bins indicates that these variations are not signiﬁcant; in other words, it is clear from the ﬁt that there was a large amount of star formation activity 8–10 Myr ago in many of these regions, but in most cases, the data do not allow us to distinguish 8 Myr old stars from 10 Myr old stars, and this non-uniqueness is reﬂected in the computed uncertainties.
Processed_The_Formation_of_Constellation_III_in_the_Large_Ma.txt	Most previous analyses of Constellation III’s SFH have concluded that the stellar arc is 10–15 Myr old. We have shown that the youngest stars in these arc struc- tures are 10–15 Myr old, but that they also contain abundant populations as old as 30 Myr. This ex- tended epoch of star formation is diﬃcult to recon- cile with currently-proposed ideas about the forma- tion mechanism of Constellation III. In the “super- supernova” or GRB shockwave scenario, star forma- tion would be triggered throughout the arc structure on a short timescale, resulting in a small age spread among the stars in Constellation III. The SSPSF sce- nario predicts a more protracted epoch of star forma- tion, but as it is usually discussed in the literature (Dopita et al. 1985; Olsen et al. 1997), the propaga- tion wave is directed radially outward from the center of Constellation III. Our SFH map conﬁrms that there are no large-scale age gradients in Constellation III that would be required by this model.
Processed_The_Formation_of_Constellation_III_in_the_Large_Ma.txt	Figure 4: The SFH map for the Constellation III region. Each panel displays the star formation rates for a single time step in the LMC’s history, from 12 Gyr ago in the upper left to 5 Myr ago in the lower right. Within a panel, the greyscale is proportional to the relative star formation rate in each grid cell (with darker color corresponding to a larger star formation rate).
Processed_The_Formation_of_Constellation_III_in_the_Large_Ma.txt	It may seem unlikely that it would be possible to sweep material into large, coherent structures without triggering massive, rapid star formation in the mate- rial. However, the LMC currently contains an exam- ple of just such a large, coherent prestellar cloud com- plex: there is a ridge of molecular gas extending more than 1.5 kpc southward from 30 Doradus, with a typ- ical width of 100 pc (Mizuno et al. 2001). This ridge contains abundant molecular and atomic gas, and yet its speciﬁc star formation rate is currently quite low. While the physical processes that led to the forma- tion of this giant molecular ridge may be quite dif- ferent from the forces that scuplted Constellation III, its existence is evidence that it is at least possible to gather prestellar material into a large coherent struc- ture, without simultaneously triggering rapid star for- mation throughout it.
Processed_The_Formation_of_Constellation_III_in_the_Large_Ma.txt	northern LMC disk. We ﬁnd that stars in the giant stellar arcs in this region formed over an extended pe- riod, from 30 to 10 Myr ago. While there are signiﬁ- cant spatial variations in the SFH, there don’t appear to be large-scale age gradients as would be expected in a SSPSF formation scenario.
Processed_The_Formation_of_Constellation_III_in_the_Large_Ma.txt	Since our detailed SFH reconstruction of Constel- lation III ﬁts neither of the widely-discussed formation scenarios for this unique structure, we propose a new scenario in which the prestellar material was swept up into large arcs, but star formation was not immedi- ately triggered throughout the cloud, or at least not violently so. The molecular ridge south of 30 Doradus provides evidence that it is possible to organize kpc- scale coherent structures in prestellar material without immediately triggering rapid star formation in the gas.
Processed_The_Formation_of_Constellation_III_in_the_Large_Ma.txt	Figure 5: The summed total SFH of the entire Constellation III region, shown as the star forma- tion rate as a function of time. The time axis is displayed linearly, but the scale changes in the three panels so that the very narrow time inter- vals at the young end can be displayed. The left panel covers about 100 Myr, the middle panel cov- ers about 1 Gyr, and the right panel covers more than 10 Gyr. The black histogram represents the SFH of the stellar arcs in the Constellation III re- gion, while the grey histogram represents the SFH of the background population in this region.
Processed_The_Formation_of_Constellation_III_in_the_Large_Ma.txt	versities for Research in Astronomy, Inc., under NASA contract NAS 5-26555. DZ acknowledges ﬁnancial sup- port from National Science Foundation grant AST- 0307482 and a fellowship from the David and Lucille Packard Foundation.
Processed_Structure,_Evolution_and_Nucleosynthesis_of_Primor.txt	contributed to some extent to the production of Mg and Al and possibly s-elements (despite the lack of iron seeds) in the early universe.
Processed_Structure,_Evolution_and_Nucleosynthesis_of_Primor.txt	The ﬁrst stars that formed in our universe, commonly referred to as Population III stars, have a composition given by the products of the big bang nucleosynthesis and are essentially made of hydrogen and helium with some traces of lithium. A major issue concerning primordial stars is the determination of the initial mass function (IMF). In the absence of heavy elements and dust grains, the cooling mechanisms may indeed be less eﬃcient and favor the formation of massive or very massive stars (e.g. Silk 1977, Bromm et al. 1999). However, it has also been shown (e.g. Carlberg 1981, Palla et al. 1983) that even a small fraction of molecular hydrogen can provide a signiﬁcant contribution to the cooling due to rotational and vibrational transitions. The resulting Jeans mass of a pure H and He cloud can then be relatively small and may even fall below 0.1 M⊙. Yoshii and Saio (1986), using the opacity-limited fragmentation theory of Silk (1977), estimate the peak mass of the IMF to be around 4-10 M⊙. Based on one-dimensional hydrodynamical simulations of a collapsing cloud, Nakamura and Umemura (1999) ﬁnd that the typical mass of the ﬁrst stars is about 3 M⊙ which may grow to 16 M⊙ by accretion. Extending their previous work by performing 2D simulations, Nakamura and Umemura (2001) re-evaluated the IMF and came up with the result that it is likely to be bimodal, with peaks around 1 and 100 M⊙. Adopting a diﬀerent approach based on the eﬀects of a primordial generation of stars on the pollution of the intergalactic medium, Abia et al. (2001) showed that an IMF peaked around 4-8 M⊙ is needed to account for the large [C,N/Fe] ratio observed in extremely metal-poor stars. Therefore, the mass scale of Pop III stars still remains relatively uncertain and it is possible that even very low-mass stars (M . 0.8 M⊙) might have survived up to present day as nuclear burning stars.
Processed_Structure,_Evolution_and_Nucleosynthesis_of_Primor.txt	In the early universe, low- and intermediate-mass stars actively contributed to the enrichment of the interstellar medium and an understanding of their evolution and nucleosynthesis is absolutely essential for a comprehension of the chemical evolution of the universe. On the theoretical side, several works have been devoted to the study of the structure and evolution of primordial stars and we refer the reader to the recent reviews by Castellani (2000) and Chiosi (2000) for a description of these previous investigations.
Processed_Structure,_Evolution_and_Nucleosynthesis_of_Primor.txt	In this paper we present and analyze computations of zero-metallicity stars, using a state- of-the-art code, thus continuing and completing the previous studies. In §2, we describe brieﬂy the physics of the stellar evolution code, and in §3 we analyze the structure and evolution of the primordial stars. We also make a careful comparison of our results with those from previous computations. In §4, we describe the evolution of the surface chemical composition in relation to the chemical history of the universe and then conclude.
Processed_Structure,_Evolution_and_Nucleosynthesis_of_Primor.txt	The Grenoble stellar evolution code has been used for these computations. For a detailed discussion of its input physics, we refer the reader to several papers published by the group (Forestini & Charbonnel 1997, Siess et al. 2000). Brieﬂy, in the domain of high temperatures (> 8000K) we use the OPAL opacity tables (Iglesias & Rogers 1996) which are adequate for enhanced values of C and O. Thus, when the C and O abundances are modiﬁed, i.e. during the third dredge-up events, their eﬀects on the opacity are taken into account.
Processed_Structure,_Evolution_and_Nucleosynthesis_of_Primor.txt	The conductive opacities are computed from a modiﬁed version of the Iben (1975) ﬁts to the Hubbard and Lampe (1969) tables for non-relativistic electrons, from Itoh et al. (1983) and Mitake et al. (1984) for relativistic electrons and from formulae of Itoh et al. (1984) as well as Raikh and Yakovlev (1982) for solid plasmas. Below 8000K, we use the atomic and molecular opacities provided by Alexander and Fergusson (1994).
Processed_Structure,_Evolution_and_Nucleosynthesis_of_Primor.txt	The equation of state (EOS) follows the formalism developed by Pols et al. (1995) and provides an accurate treatment of the non-ideal eﬀects, especially the treatment of pressure ionization and degeneracy in partially ionized plasma (see e.g. Siess et al. 2000 for a detailed discussion and for the tests of this EOS).
Processed_Structure,_Evolution_and_Nucleosynthesis_of_Primor.txt	The nuclear network includes 180 nuclear (neutron, proton and 4He captures) and decay re- actions and we follow the abundance evolution of 52 nuclides. Most of the reaction rates which originally came from Caughlan and Fowler (1988) have recently been updated using the NACRE compilation (Angulo et al. 1999). Finally, the nuclear screening factors are parameterized using the Graboske et al. (1973) formalism, including weak, intermediate and strong screening cases and the computations are stopped in the massive stars (M & 10 M⊙) at C-ignition when the temperature reaches 8 × 108K.
Processed_Structure,_Evolution_and_Nucleosynthesis_of_Primor.txt	The surface boundary conditions are treated in the grey atmosphere approximation and we use the de Jager mass loss rate prescription (de Jager et al. 1988). However, because of the lack of metals, the mass loss rate only activates during the late stages of the AGB phase.
Processed_Structure,_Evolution_and_Nucleosynthesis_of_Primor.txt	We use the standard mixing length theory (Cox & Guili 1968) with αMLT = 1.5 and we follow the Schaller et al. (1992) suggestion and account for a moderate overshooting characterized by d = 0.20HP during the central H and He burning phases. During the AGB phase, the extent of the convective envelope has also been increased by assuming a moderate overshooting characterized by d = 0.05HP . This choice was made as a way of obtaining a stable Schwarzschild boundary for the envelope in a simple and consistent way, and is similar to the work of Herwig et al. (1999).
Processed_Structure,_Evolution_and_Nucleosynthesis_of_Primor.txt	The chemical composition of our initial models is characterized by X = 0.765, Y = 0.235 and by a 7Li mass fraction XLi = 9.26 × 10−10. These values (Burles & Tytler 1998, Bonifacio & Molaro 1997) are taken from Big Bang nucleosynthesis calculations. Finally, inside convective zones, the chemical species are homogeneously mixed.
Processed_Structure,_Evolution_and_Nucleosynthesis_of_Primor.txt	For numerical details, the code uses an adaptive mesh point. Diﬀerent criteria regulate the zoning and in particular, shells are added in regions where the gradients of energy production and chemical composition are steep. Typically, the region of mixing episodes is described by ≈ 500 shells and depending on the evolutionary status, the structure is discretized into 1000-3500 shells. The time step is constrained so that the relative variations of the independent variables, between one model to the next, do not exceed 10% and the luminosity does not diﬀer by more that 5%. During the convergence process, the equations of stellar structure must be satisﬁed at least to a 10−3 level of precision.
Processed_Structure,_Evolution_and_Nucleosynthesis_of_Primor.txt	In this section, we describe the structure and evolution of low- and intermediate-mass stars through the computations of 1 and 5 M⊙ models. These models were chosen so as to enable comparison with other works. Firstly, we analyze the evolution during the core H and He burning, then we describe the AGB phase and ﬁnally we summarize the nucleosynthesis occurring in these stars.
Processed_Structure,_Evolution_and_Nucleosynthesis_of_Primor.txt	During the pre-main sequence (PMS) contraction phase, the small amount of primordial 7Li is burnt by proton-captures to helium. After 12.5 Myr5, the central temperature reaches ≃ 107K and a convective core of ≃ 0.20 M⊙ forms. There H is burnt mainly via by the PPI (97%) chain which is later supplemented by the PPIII chain. After 3.72 × 109 yr the convective core disappears, and due to the increased temperature, carbon production by the 3α reaction starts. When the central carbon mass fraction reaches 1.8 × 10−12 (Tc = 6.37 × 107K), the CN cycle ultimately ignites.
Processed_Structure,_Evolution_and_Nucleosynthesis_of_Primor.txt	The shifting of the nuclear energy production from the p-p chain to the CNO cycle takes place roughly at the same carbon abundance threshold. This is illustrated in Fig. 4.1 which depicts the evolution of the central carbon abundance as a function of central H abundance. The curves show a bump around a similar value of log X(12C) ≃ −11.5. This indicates an increase in the H mass fraction, which results from the development of the convective core, corresponding to the ignition of the CNO reactions. We can see that stars with M ≥ 3 M⊙ ignite the CNO cycle relatively early in the evolution, when the central H abundance X ≥ 0.1. On the other hand, stars less massive that ∼ 1 M⊙ never ignite CNO reactions during their main sequence.

Subsets and Splits

No community queries yet

The top public SQL queries from the community will appear here once available.