Multiple Imputation Methods Research Articles

Data Missing Not at Random in Mobile Health Research: Assessment of the Problem and a Case for Sensitivity Analyses.

BackgroundMissing data are common in mobile health (mHealth) research. There has been little systematic investigation of how missingness is handled statistically in mHealth randomized controlled trials (RCTs). Although some missing data patterns (ie, missing at random [MAR]) may be adequately addressed using modern missing data methods such as multiple imputation and maximum likelihood techniques, these methods do not address bias when data are missing not at random (MNAR). It is typically not possible to determine whether the missing data are MAR. However, higher attrition in active (ie, intervention) versus passive (ie, waitlist or no treatment) conditions in mHealth RCTs raise a strong likelihood of MNAR, such as if active participants who benefit less from the intervention are more likely to drop out.ObjectiveThis study aims to systematically evaluate differential attrition and methods used for handling missingness in a sample of mHealth RCTs comparing active and passive control conditions. We also aim to illustrate a modern model-based sensitivity analysis and a simpler fixed-value replacement approach that can be used to evaluate the influence of MNAR.MethodsWe reanalyzed attrition rates and predictors of differential attrition in a sample of 36 mHealth RCTs drawn from a recent meta-analysis of smartphone-based mental health interventions. We systematically evaluated the design features related to missingness and its handling. Data from a recent mHealth RCT were used to illustrate 2 sensitivity analysis approaches (pattern-mixture model and fixed-value replacement approach).ResultsAttrition in active conditions was, on average, roughly twice that of passive controls. Differential attrition was higher in larger studies and was associated with the use of MAR-based multiple imputation or maximum likelihood methods. Half of the studies (18/36, 50%) used these modern missing data techniques. None of the 36 mHealth RCTs reviewed conducted a sensitivity analysis to evaluate the possible consequences of data MNAR. A pattern-mixture model and fixed-value replacement sensitivity analysis approaches were introduced. Results from a recent mHealth RCT were shown to be robust to missing data, reflecting worse outcomes in missing versus nonmissing scores in some but not all scenarios. A review of such scenarios helps to qualify the observations of significant treatment effects.ConclusionsMNAR data because of differential attrition are likely in mHealth RCTs using passive controls. Sensitivity analyses are recommended to allow researchers to assess the potential impact of MNAR on trial results.

Comparison of methods for filling in gaps in monthly rainfall series in arid regions

AbstractFull and reliable rainfall data are required for hydrological modelling, integrated water resource management, and planning. However, these data suffer from record gaps and sparse rain gage distribution, which implies that the use of an imputation method is crucial. This study aims to compare the outputs from 10 imputation methods that were used to infill missing rainfall depth data (MRD) in an arid Mediterranean region. Different statistical tests were used to assess the outputs from the imputation methods. The results showed that for MRD ranges between 5 and 20% the stepwise multiple linear regression (MLRsw) method was valid and produced the best results with a root‐mean‐square error (RSME) and mean absolute error (MAE) of less than 7 and 2 mm, respectively. This was followed by the Monte Carlo Markov chain expectation–maximization‐based multiple imputation (MI‐MCMC) method, which had an RSME and MAE of 1.01 and 0.08 mm, respectively, at 20% MRD. On the other hand, the use of satellite data for imputation (LR_GPCC estimates) was appropriate for MRD ranging between 10 and 15%, while the statistical and spatial method was suitable for MRD of less than 5%.

Estimating the Influence of Incarceration on Subsequent Experience With Violence Among Black Men Who Have Sex With Men in the HPTN061 Study.

Black men who have sex with men (BMSM) are disproportionately incarcerated in the United States. Incarceration is a barrier to health equity and may be a risk factor for experiences of interpersonal violence. However, the effect of incarceration on experienced violence among BMSM is understudied. We examined associations between recent incarceration on subsequent experiences of race- or sexuality-based violence, intimate partner violence, or community violence. We analyzed data from the HPTN 061 study. Analysis includes data on 1,169 BMSM recruited from 6 U.S. cities who were present at baseline as well as 6- and 12-month follow-up interview. We tested if self-reported incarceration between baseline and 6 months was associated with self-reported outcomes between 6 and 12 months using logistic regression with inverse probability of treatment weighting and multiple imputation methods. Experienced outcomes included violence due to race or sexuality, intimate partner violence and aggression, and community violence (i.e., gang violence, robbery, shooting). Approximately 14% reported incarceration between baseline and 6 months and 90% reported experiencing violence between 6 and 12 months. In adjusted analyses, incarceration was associated with subsequent race- or sexuality-based violence [aOR (adjusted odds ratio) range: 1.25-1.41, 95% CI (confidence interval) range: 1.00-1.74], experiences of physical abuse and aggression from intimate partners (aOR: 2.35; 95% CI: 1.50, 3.70) and community violence (OR 1.82; 95% CI: 1.23, 2.72). Recent incarceration experience increased risk of exposure to future violence in this population. Mixed methods research examining mediating paths between and downstream effects of incarceration and violence on the wellbeing and health of BMSM is needed. We implore researchers to study violence and incarceration among BMSM. Practitione should implement strategies such as trauma-informed interventions, and policies strengthening the social and economic support needs of Black populations.

P375 A simulation study to evaluate the performance of a multiple imputation method to address missing data in an analysis of clinical effectiveness using the ImproveCareNow Registry of pediatric patients with Crohn’s disease

Abstract Background Patient registries are a source of real-world data (RWD) that are increasingly used to generate real-world evidence (RWE) for the safety and effectiveness of medical therapies. However, in registries, outcome data typically collected in clinical trials may be missing in some patients, creating a major challenge in evaluating RWD reliably. Clinical remission in pediatric Crohn’s disease (PCD), defined by the Short Pediatric Crohn’s disease Activity Index (sPCDAI) <10, is routinely collected in the ImproveCareNow (ICN) Registry, the world’s largest registry of pediatric patients with inflammatory bowel disease (IBD). A feasibility analysis of ICN as a RWD source to estimate the efficacy of therapies in PCD found about one-third of patients were missing at least 1 of the 6 components needed to calculate the sPCDAI scores at Week 52. We conducted a simulation study to evaluate the performance of the multiple imputation (MI) method in addressing the missing data issue. Methods A Starting Dataset from the ICN registry included CD patients age 2 to <18 yr when beginning treatment with a biologic agent other than ustekinumab during 2014 to 2019, who had a baseline visit and a “Week 52” visit. A Complete Dataset included only patients in the Starting Dataset who had all components required to calculate a sPCDAI score at the Week 52 visit. The true remission rate was calculated for the Complete Dataset. For the simulation, missing data patterns for individual sPCDAI components that were observed in the Starting Dataset were randomly imposed in the Complete Dataset. Variables that were predictive of sPCDAI scores, its components, and sPCDAI missingness were included. Ten thousand datasets were simulated for each of three scenarios: base case of the observed missingness in the Starting Dataset, 1.5 times the base case and 2 times the base case. The MI method was applied to the simulated datasets to estimate the remission rate. MI performance was measured by bias and coverage of the true remission rate from the Complete Dataset. Results The Starting and Complete Datasets included 1,458 and 1,212 patients, respectively. The true clinical remission rate in the Complete Dataset was 75.1% (95% CI: 72.6%-77.5%). Analysis without MI for the missing sPCDAI data resulted in underestimation of clinical remission rate, with bias about -2% to -5% (Table 1). The MI method eliminated the bias and had 100% coverage in all simulation scenarios by using a large number of independent predictors. Conclusion The MI method improved the validity, reliability, and efficiency of using RWD for estimating clinical remission in pediatric patients with CD and could be valuable in other registry studies.

Multiple imputation of missing data in multilevel models with the R package mdmb: a flexible sequential modeling approach

Multilevel models often include nonlinear effects, such as random slopes or interaction effects. The estimation of these models can be difficult when the underlying variables contain missing data. Although several methods for handling missing data such as multiple imputation (MI) can be used with multilevel data, conventional methods for multilevel MI often do not properly take the nonlinear associations between the variables into account. In the present paper, we propose a sequential modeling approach based on Bayesian estimation techniques that can be used to handle missing data in a variety of multilevel models that involve nonlinear effects. The main idea of this approach is to decompose the joint distribution of the data into several parts that correspond to the outcome and explanatory variables in the intended analysis, thus generating imputations in a manner that is compatible with the substantive analysis model. In three simulation studies, we evaluate the sequential modeling approach and compare it with conventional as well as other substantive-model-compatible approaches to multilevel MI. We implemented the sequential modeling approach in the R package mdmb and provide a worked example to illustrate its application.

Evaluation of Four Multiple Imputation Methods for Handling Missing Binary Outcome Data in the Presence of an Interaction between a Dummy and a Continuous Variable

Multiple imputation by chained equations (MICE) is the most common method for imputing missing data. In the MICE algorithm, imputation can be performed using a variety of parametric and nonparametric methods. The default setting in the implementation of MICE is for imputation models to include variables as linear terms only with no interactions, but omission of interaction terms may lead to biased results. It is investigated, using simulated and real datasets, whether recursive partitioning creates appropriate variability between imputations and unbiased parameter estimates with appropriate confidence intervals. We compared four multiple imputation (MI) methods on a real and a simulated dataset. MI methods included using predictive mean matching with an interaction term in the imputation model in MICE (MICE-interaction), classification and regression tree (CART) for specifying the imputation model in MICE (MICE-CART), the implementation of random forest (RF) in MICE (MICE-RF), and MICE-Stratified method. We first selected secondary data and devised an experimental design that consisted of 40 scenarios (2 × 5 × 4), which differed by the rate of simulated missing data (10%, 20%, 30%, 40%, and 50%), the missing mechanism (MAR and MCAR), and imputation method (MICE-Interaction, MICE-CART, MICE-RF, and MICE-Stratified). First, we randomly drew 700 observations with replacement 300 times, and then the missing data were created. The evaluation was based on raw bias (RB) as well as five other measurements that were averaged over the repetitions. Next, in a simulation study, we generated data 1000 times with a sample size of 700. Then, we created missing data for each dataset once. For all scenarios, the same criteria were used as for real data to evaluate the performance of methods in the simulation study. It is concluded that, when there is an interaction effect between a dummy and a continuous predictor, substantial gains are possible by using recursive partitioning for imputation compared to parametric methods, and also, the MICE-Interaction method is always more efficient and convenient to preserve interaction effects than the other methods.

Asymptotic Standard Errors of Generalized Partial Credit Model True Score Equating Using Characteristic Curve Methods.

In this study, the delta method was applied to estimate the standard errors of the true score equating when using the characteristic curve methods with the generalized partial credit model in test equating under the context of the common-item nonequivalent groups equating design. Simulation studies were further conducted to compare the performance of the delta method with that of the bootstrap method and the multiple imputation method. The results indicated that the standard errors produced by the delta method were very close to the criterion empirical standard errors as well as those yielded by the bootstrap method and the multiple imputation method under all the manipulated conditions.

Multiple imputation using nearest neighbor methods

Missing values are a major problem in medical research. As the complete case analysis discards useful information, estimation and inference may suffer strongly. Multiple imputation has been shown to be a useful strategy to handle missing data problems and account for the uncertainty of imputation. In the presence of high-dimensional data (p≫n), the missing values raise even more serious problems as the existing software packages tend to fail. We present multiple imputation methods based on nearest neighbors. The distances are computed using the information of correlation among the target and candidate predictors. Thus only the relevant predictors contribute for computing distances. The method successfully imputes missing values also in high-dimensional settings. Using a variety of simulated data with MCAR and MAR missing patterns, the proposed algorithm is compared to existing methods. Various measures are used to compare the performance of methods, including MSE for imputation, MSE of estimated regression coefficients, their standard errors, confidence intervals, and their coverage probabilities. The simulation results, for both cases n<p and n>p, show that the sequential imputation using weighted nearest neighbors can be successfully applied to a wide range of data settings and outperforms or is close to the best when compared to existing methods.

Comparison of Missing Rainfall Data Treatment Analysis at Kenyir Lake

Rainfall is one of the frequent data used in weather-related studies. Sometimes the data have missing information that needs the treatment to make sure the data can be useful, complete and reliable. There are many methods in treating missing data suggested by previous studies. The best selected method to estimate missing rainfall data in different regions may vary depending on the rainfall pattern and spatial distribution. Therefore, this paper discussed and compared 3 different methods in missing data treatment. The selected methods are Expectation Maximization (EM), Inverse Distance Weighted (IDW) and Multiple Imputation (MI). After analysis, the best method is IDW based on root mean square error (RMSE), mean absolute error (MAE), correlation coefficient (r) and percentage of error (% of error) values. The IDW method has RMSE, MAE values and the lowest % of error values. In addition, the r value of IDW method is highest compared to EM and MI method. MI method recorded the highest values of RMSE, MAE and % of error with the lowest r value that proved MI method is the least accurate method to use in missing data treatment. After all methods were implemented, it proved that the IDW method is the best way to treat missing data because the analysis shows monthly rainfall distribution for 4 treatment stations in line to 3 missing data stations compared to EM and MI methods.

Comparison of Single and MICE Imputation Methods for Missing Values: A Simulation Study

High quality data is essential in every field of research for valid research findings. The presence of missing data in a dataset is common and occurs for a variety of reasons such as incomplete responses, equipment malfunction and data entry error. Single and multiple data imputation methods have been developed for data imputation of missing values. This study investigated the performance of single imputation using mean and multiple imputation method using Multivariate Imputation by Chained Equations (MICE) via a simulation study. The MCAR which means missing completely at random were generated randomly for ten levels of missing rates (proportion of missing data): 5% to 50% for different sample sizes. Mean Square Error (MSE) was used to evaluate the performance of the imputation methods. Data imputation method depends on data types. Mean imputation is commonly used to impute missing values for continuous variable while MICE method can handle both continuous and categorical variables. The simulation results indicate that group mean imputation (GMI) performed better compared to overall mean imputation (OMI) and MICE with lowest value of MSE for all sample sizes and missing rates. The MSE of OMI, GMI, and MICE increases when missing rate increases. The MICE method has the lowest performance (i.e. highest MSE) when percentage of missing rates is more than 15%. Overall, GMI is more superior compared to OMI and MICE for all missing rates and sample size for MCAR mechanism. An application to a real dataset confirmed the findings of the simulation results. The findings of this study can provide knowledge to researchers and practitioners on which imputation method is more suitable when the data involves missing data.

Applications of Regression based multiple imputation methods considering logical boundary information of missing values in financial and health survey data

Applications of Regression based multiple imputation methods considering logical boundary information of missing values in financial and health survey data

Using multiple imputation to classify potential outcomes subgroups.

With medical tests becoming increasingly available, concerns about over-testing, over-treatment and health care cost dramatically increase. Hence, it is important to understand the influence of testing on treatment selection in general practice. Most statistical methods focus on average effects of testing on treatment decisions. However, this may be ill-advised, particularly for patient subgroups that tend not to benefit from such tests. Furthermore, missing data are common, representing large and often unaddressed threats to the validity of most statistical methods. Finally, it is often desirable to conduct analyses that can be interpreted causally. Using the Rubin Causal Model framework, we propose to classify patients into four potential outcomes subgroups, defined by whether or not a patient's treatment selection is changed by the test result and by the direction of how the test result changes treatment selection. This subgroup classification naturally captures the differential influence of medical testing on treatment selections for different patients, which can suggest targets to improve the utilization of medical tests. We can then examine patient characteristics associated with patient potential outcomes subgroup memberships. We used multiple imputation methods to simultaneously impute the missing potential outcomes as well as regular missing values. This approach can also provide estimates of many traditional causal quantities of interest. We find that explicitly incorporating causal inference assumptions into the multiple imputation process can improve the precision for some causal estimates of interest. We also find that bias can occur when the potential outcomes conditional independence assumption is violated; sensitivity analyses are proposed to assess the impact of this violation. We applied the proposed methods to examine the influence of 21-gene assay, the most commonly used genomic test in the United States, on chemotherapy selection among breast cancer patients.

An efficient ensemble method for missing value imputation in microarray gene expression data

BackgroundThe genomics data analysis has been widely used to study disease genes and drug targets. However, the existence of missing values in genomics datasets poses a significant problem, which severely hinders the use of genomics data. Current imputation methods based on a single learner often explores less known genomic data information for imputation and thus causes the imputation performance loss.ResultsIn this study, multiple single imputation methods are combined into an imputation method by ensemble learning. In the ensemble method, the bootstrap sampling is applied for predictions of missing values by each component method, and these predictions are weighted and summed to produce the final prediction. The optimal weights are learned from known gene data in the sense of minimizing a cost function about the imputation error. And the expression of the optimal weights is derived in closed form. Additionally, the performance of the ensemble method is analytically investigated, in terms of the sum of squared regression errors. The proposed method is simulated on several typical genomic datasets and compared with the state-of-the-art imputation methods at different noise levels, sample sizes and data missing rates. Experimental results show that the proposed method achieves the improved imputation performance in terms of the imputation accuracy, robustness and generalization.ConclusionThe ensemble method possesses the superior imputation performance since it can make use of known data information more efficiently for missing data imputation by integrating diverse imputation methods and learning the integration weights in a data-driven way.

A pharmacoepidemiological study of the association of suicide reattempt risk with psychotropic drug exposure

IntroductionRecent pharmacoepidemiological studies have suggested that consumption of certain classes of psychotropic drugs could be considered protective or risk factors for suicidal behaviour. The aim of the study was to evaluate the association between the risk of suicide reattempt within 6 and 14 months after a suicide attempt (SA) with the use of different classes of psychotropic drugs, combination pairs and treatment adequacy from inclusion through 6 and 14 months post-SA. MethodA prospective observational cohort of 972 subjects from the ALGOS study from January 2010 to February 2013 was used to evaluate the association of risk of suicide reattempt within 6 and 14 months with the use of different classes of psychotropic drugs (antidepressants, anxiolytics, antipsychotics, lithium, anticonvulsants, analgesics, opioid maintenance therapy and maintenance treatment for alcohol dependence). A multivariable Cox model was performed after imputation of missing data using the multiple imputation method. ResultsOur main results did not show an association between psychotropic drug use and suicide reattempt after 6 months of follow-up. We demonstrated that the use of benzodiazepines (HR = 1.87 [1.25; 2.81], p < 0.01) and hypnotics (HR = 1.49 [1.03; 2.17], p = 0.04) or a combination of both (HR = 1.80 [1.17; 2.72], p = 0.01) were associated with suicide reattempt within 14 months after a previous SA. ConclusionThe early identification of a positive association between psychotropic drugs and the risk of suicidal behaviour is extremely important for prevention of suicide reattempts. Special precautions should be considered when prescribing psychotropic drugs for these subjects, particularly those at risk of suicide reattempt.

Pirfenidone in patients with progressive fibrotic interstitial lung diseases other than idiopathic pulmonary fibrosis (RELIEF): a double-blind, randomised, placebo-controlled, phase 2b trial

Pirfenidone has been shown to slow disease progression in patients with idiopathic pulmonary fibrosis (IPF). However, there are few treatment options for progressive fibrotic interstitial lung diseases (ILDs)) other than IPF. In view of the pathomechanistic and clinical similarities between IPF and other progressive fibrotic ILDs, we aimed to assess the efficacy and safety of pirfenidone in patients with four non-IPF progressive fibrotic ILDs. We did a multicentre, double-blind, randomised, placebo-controlled, parallel phase 2b trial (RELIEF) in 17 centres with expertise in ILD in Germany. Eligible participants were patients aged 18-80 years with progressive fibrotic ILD due to four diagnoses: collagen or vascular diseases (ie, connective tissue disease-associated ILDs), fibrotic non-specific interstitial pneumonia, chronic hypersensitivity pneumonitis, or asbestos-induced lung fibrosis. Other eligibility criteria included a forced vital capacity (FVC) of 40-90% predicted, a diffusing capacity of the lung for carbon monoxide of 10-90% predicted, and an annual decline of FVC of at least 5% predicted despite conventional therapy, based on at least three measurements within 6-24 months before enrolment. Patients who had received any previous antifibrotic therapy were excluded. We randomly assigned patients (1:1) to either oral pirfenidone (267 mg three times per day in week 1, 534 mg three times per day in week 2, and 801 mg three times per day thereafter) or matched placebo, added to their ongoing medication. Randomisation was done centrally using permuted block randomisation with varying block sizes stratified by the four diagnostic groups. Patients, investigators, statisticians, monitors, and the study coordinator were masked to treatment assignment until database closure. The placebo-controlled study period was 48 weeks (including up-titration). The primary endpoint was absolute change in percentage of predicted FVC (FVC % predicted) from baseline to week 48 in the intention-to-treat population, with imputation of missing data by the smallest sum of squared differences and attribution of deceased patients to the lowest rank in a rank ANCOVA model. Additionally, we did linear mixed-model repeated measures slope analyses of FVC % predicted longitudinal data over the course of the study as a prespecified sensitivity analysis and post-hoc sensitivity analyses of the primary endpoint in the intention-to-treat population using imputation methods of last observation carried forward [LOCF] and a regression-based multiple imputation procedure. Safety was assessed in all patients who received at least one dose of study medication. This trial is registered with EudraCT 2014-000861-32; DRKS00009822 and is no longer recruiting. Between April 5, 2016, and Oct 4, 2018, we randomly assigned 127 patients to treatment: 64 to pirfenidone, 63 to placebo. After 127 patients had been randomised, the study was prematurely terminated on the basis of an interim analysis for futility triggered by slow recruitment. After 48 weeks and in the overall population of 127 patients, rank ANCOVA with diagnostic group included as a factor showed a significantly lower decline in FVC % predicted in the pirfenidone group compared with placebo (p=0·043); the result was similar when the model was stratified by diagnostic group (p=0·042). A significant treatment effect was also observed when applying the LOCF and multiple imputation methods to analyses of the primary endpoint. The median difference (Hodges-Lehmann estimate) between pirfenidone and placebo groups for the primary endpoint was 1·69 FVC % predicted (95% CI -0·65 to 4·03). In the linear mixed-model repeated measures slope analysis of FVC % predicted, the estimated difference between treatment and placebo groups from baseline to week 48 was 3·53 FVC % predicted (95% CI 0·21 to 6·86) with imputation of deaths as prespecified, or 2·79 FVC % predicted (95% CI 0·03 to 5·54) without imputation. One death (non-respiratory) occurred in the pirfenidone group (2%) and five deaths (three of which were respiratory) occurred in the placebo group (8%). The most frequent serious adverse events in both groups were infections and infestations (five [8%] in the pirfenidone group, ten [16%] in the placebo group); general disorders including disease worsening (two [3%] in the pirfenidone group, seven [11%] in the placebo group); and cardiac disorders (one ([2%] in the pirfenidone group, 5 [8%] in the placebo group). Adverse events (grade 3-4) of nausea (two patients on pirfenidone, two on placebo), dyspnoea (one patient on pirfenidone, one on placebo), and diarrhoea (one patient on pirfenidone) were also observed. In view of the premature study termination, results should be interpreted with care. Nevertheless, our data suggest that in patients with fibrotic ILDs other than IPF who deteriorate despite conventional therapy, adding pirfenidone to existing treatment might attenuate disease progression as measured by decline in FVC. German Center for Lung Research, Roche Pharma.

Roles of effort and reward in well-being for police officers in Singapore: The effort-reward imbalance model

RationaleDespite the growing body of research on the effort-reward imbalance (ERI) model, evidence is scarce for the generalizability of the model across various national and occupational contexts. Also, studies that examine a wider variety of health and well-being outcomes of ERI are warranted, especially in vocations in which effort and stress is known to be high and reward is usually limited – such as policing. ObjectiveThe current study examined the relationship between the ERI perception and four employee well-being outcomes (self-reported physical health, sickness absence, job satisfaction, and work-life effectiveness) among a sample of police officers in Singapore. MethodsCross-sectional survey data collected from a large representative sample of uniformed police officers from Singapore were used (N = 8729). Missing data were handled with the multiple imputation method and logistic regression analyses were used to test hypotheses. ResultsIn line with the theory, employees characterized by high intrinsic effort and low reward (ERI > 1) reported significantly elevated odds ratios of poor physical health (OR = 1.25), job dissatisfaction (OR = 1.53), and work-life ineffectiveness (OR = 1.31). Contrary to expectations, an unusual relationship was observed such that police officers who were overcommitted exhibited lower odds ratios of the suboptimal outcomes. Interestingly, police officers in the low effort–low reward condition exhibited the worst outcomes, whereas officers in the high effort–high reward condition reported optimal outcomes. Findings were generally consistent across genders and age groups. ConclusionsThis research bolsters the core proposition of the ERI model and emphasizes the importance of social and occupational contexts in the study of ERI.

Direct and Unbiased Multiple Imputation Methods for Missing Values of Categorical Variables

Direct and Unbiased Multiple Imputation Methods for Missing Values of Categorical Variables

Prediction of Breast Cancer Survival by Machine Learning Methods: An Application of Multiple Imputation.

Background:The low breast cancer survival rates in less developed countries are critical. The machine learning techniques predict cancers survival with high accuracy. Missing data are the most important limitation for using the highest potential of these techniques to predict cancers survival. Multiple imputation (MI) was implemented and analyzed in detail to impute the missing data of a breast cancer dataset.Methods:The dataset was from The Omid Treatment and Research Center Urmia, Iran between Jan 2006 and Dec 2012 and had information from 856 women. The algorithms such as C5 and repeated incremental pruning to produce error reduction were applied on the imputed versions of the original dataset and the non-imputed dataset to predict and extract clinical rules, respectively.Results:The findings showed the performance of C5 in all the evaluation criteria including accuracy (84.42%), sensitivity (92.21%), specificity (64%), Kappa statistic (59.06%), and the area under the receiver operator characteristic (ROC) curve (0.84), was improved after imputation.Conclusion:The dataset of the present study met the requirements for using the multiple imputation method. The extracted rules after the application of MI were more comprehensive and contained knowledge that is more clinical. However, the clinical value of the extracted rules after filling in the missing data did not noticeably increase.

Trend analysis of seasonal rainfall patterns in Peninsular Malaysia

Long records of rainfall data with good quality is a significant component that triggers the entire chain of responses in a hydrological cycle by providing useful information to hydrologist and related parties. In this study, the temporal trend analysis of seasonal rainfall data was carried out in Peninsular Malaysia using the non-parametric Mann-Kendall test. Then, the magnitude of the seasonal rainfall trend line was determined using Sen’s Slope Estimator. Historical rainfall data with a minimum duration of 25 years were checked using four homogeneity tests and the missing values in the data were infilled using multiple imputation method. According to the output of the homogeneity tests, 64.3%, 7.1% and 28.6% of the rainfall time series can be categorised as useful, doubtful and suspect, respectively. From the trend test results, almost all the rainfall data were detected to have an increasing trend except Bayan Lepas for Inter-Monsoon 1, Southwest Monsoon and Inter-Monsoon 2. In addition to that, Bayan Lepas, Hospital Seremban, Ipoh and KLIA Sepang were detected to have a significant increasing trend for Northeast Monsoon as well as Batu Pahat and KLIA Sepang for Inter-Monsoon 2. In conclusion, Peninsular Malaysia exhibited increasing rainfall trend from year 1984 to 2019.

Modern multiple imputation with functional data

This work considers the problem of fitting functional models with sparsely and irregularly sampled functional data. It overcomes the limitations of the state‐of‐the‐art methods, which face major challenges in the fitting of more complex non‐linear models. Currently, many of these models cannot be consistently estimated unless the number of observed points per curve grows sufficiently quickly with the sample size, whereas we show numerically that a modified approach with more modern multiple imputation methods can produce better estimates in general. We also propose a new imputation approach that combines the ideas of MissForest with Local Linear Forest and compare their performance with PACE and several other multivariate multiple imputation methods. This work is motivated by a longitudinal study on smoking cessation, in which the electronic health records (EHR) from Penn State PaTH to Health allow for the collection of a great deal of data, with highly variable sampling. To illustrate our approach, we explore the relation between relapse and diastolic blood pressure. We also consider a variety of simulation schemes with varying levels of sparsity to validate our methods.