Time Series Bias Research Articles

Challenges and Prospects to Time Series Burst Overlap Interferometry (BOI): Some Insights From a New BOI Algorithm Test Over the Chaman Fault

How to obtain millimeter-scale along-track deformations using phase measurements is still a pending question for InSAR community. Although Burst Overlap Interferometry (BOI) technique makes this question seem tractable, most applications of BOI still focus on extracting centimeter-scale deformations for co-seismic cases. To further improve measurement accuracy, here we propose a new time series BOI algorithm towards maximizing the performance of BOI and obtaining as high accuracy of the along-track deformations as possible. This algorithm has three major steps. The first step is to enhance BOI phase signal-to-noise ratio using our newly proposed phase estimator. In the second step, we apply a strain model-based method to further suppress the phase noise and rescue more data points. In the third step, a misregistration correction procedure which considers plate motion is applied to mitigate BOI time series bias. We tested our proposed algorithm over the Chaman fault. Although the derived millimeter-scale deformations demonstrate the effectiveness of our method, experimental results show that decorrelation and ionospheric disturbance are still two great challenges of BOI techniques.

Evaluation gridded precipitation datasets in Indonesia

This preliminary study evaluates ten gridded precipitation datasets in Indonesia, namely APHRODITE, CMORPH, CHIRPS, GFD, SA-OBS, TMPA 3B42 v7, PERSIAN-CDR at 0.25°, moreover GSMaP_NRT V06, GPM-IMERG (Early-Run) V06, and MSWEP V2 at 0.1» in the period of 2003 to 2015. The evaluation focuses on time series bias using metrics such as Mean Error, Coefficient of Variation, Relative Change (Variability), and Kolmogorov-Smirnov test (KS-test) at daily, monthly, seasonal, and annual time scales. The statistical relationship between the precipitation datasets with reference observational data use Taylor diagrams for evaluating the relative skill of the precipitation dataset. The study aims to evaluate the uncertainty of the precipitation datasets compared to rain gauge datasets. Time series bias of SA-OBS and MSWEP have the nearest value to zero as the best score. The relative skill of monthly rainfall based on rainfall typical shows that MSWEP outperformed in regions A and B, GPM-IMERG in C region. GPM-IMERG's relative skill is outperformed than other datasets at annual time scale in Region A and B, while TMPA 3B42 in Region C. The application of existing precipitation datasets is essential to cope with the limitation of rain gauge observations. This study implicates the development of precipitation products in the Indonesia region.

Estimating Dynamic Common-Correlated Effects in Stata

In this article, I introduce a new command, xtdcce2, that fits a dynamic common-correlated effects model with heterogeneous coefficients in a panel with a large number of observations over cross-sectional units and time periods. The estimation procedure mainly follows Chudik and Pesaran (2015b, Journal of Econometrics 188: 393–420) but additionally supports the common correlated effects estimator (Pesaran, 2006, Econometrica 74: 967–1012), the mean group estimator (Pesaran and Smith, 1995, Journal of Econometrics 68: 79–113), and the pooled mean group estimator (Pesaran, Shin, and Smith, 1999, Journal of the American Statistical Association, 94: 621–634). xtdcce2 allows heterogeneous or homogeneous coefficients and supports instrumental-variable regressions and unbalanced panels. The cross-sectional dependence test is automatically calculated and presented in the estimation output. Small-sample time-series bias can be corrected by “half-panel” jackknife correction or recursive mean adjustment. I carry out a simulation to prove the estimator's consistency.

How do marine closures affect the analysis of catch and effort data?

Fishery managers increasingly use marine closures as a tool to conserve ecosystems, biodiversity, and fish abundance. Despite the suggested benefits of closed areas, the limited or no data collection within them leads to difficulties assessing the population status. We investigated how spatial closures impacted the reliability of indices of abundance obtained from standardization methods applied to catch per unit effort data. The presence of closed areas generally introduced a bias in the derived index of abundance, and the magnitude of bias increased as the portion of the population in closed areas increased. In general, restricting the data to the areas that have been continuously fished over time performed best when spatial closures protected a small to medium portion of the population. However, as the portion of the population that was protected increased, the time series bias associated with this approach increased, and the use of an imputation approach was needed for adequate performance. Similarly, the collection of ancillary data in the closed area reduced bias in the estimate of final year depletion when area closures protected a large portion of the population.

Four decades of glacier mass balance observations in the Indian Himalaya

Understanding the glacier mass balance is necessary to explain the rate of shrinkage and to infer the impact of climate change. The present study provides an overview of the glacier mass balance records by glaciological, geodetic, hydrological and accumulation-area ratio (AAR) and specific mass balance relationship methods in the Indian Himalaya since 1970s. It suggests that the mass balance measurements by glaciological methods have been conducted for ten glaciers in the western Himalaya, four glaciers in the central Himalaya and one in the eastern Himalaya. Hydrological mass balance has been conducted only on Siachen Glacier from 1987 to 1991. Geodetic method has been attempted for the Lahaul–Spiti region for a short time span during 1999–2011 and Hindu Kush–Karakoram–Himalaya region from 2003 to 2008. We compared in situ specific balance data series with specific mass balance derived from AAR and specific mass balance relationship. The results derived from existing and newly presented regression model based on AAR and specific mass balance relationship induced unrealistic specific mass balance for several glaciers. We also revised AAR0 and ELA0 based on available in situ AAR and specific mass balance data series of Indian Himalayan glaciers. In general, in situ specific and cumulative specific mass balance observed over different regions of the Indian Himalayan glaciers shows mostly negative mass balance years with a few positive ones during 1974–2012. On a regional level, the geodetic studies suggest that on the whole western, the central and the eastern Himalaya experienced vast thinning during the last decade (2000s). Conversely, Karakoram region showed slight mass gain during almost similar period. However, the glaciological, hydrological and geodetic mass balance data appear to exhibit short time series bias. We therefore recommend creation of benchmark glaciers network for future research to determine the impact of climate change on the Himalayan cryosphere.

Performance Metrics for Soil Moisture Downscaling Methods: Application to DISPATCH Data in Central Morocco

Data disaggregation (or downscaling) is becoming a recognized modeling framework to improve the spatial resolution of available surface soil moisture satellite products. However, depending on the quality of the scale change modeling and on the uncertainty in its input data, disaggregation may improve or degrade soil moisture information at high resolution. Hence, defining a relevant metric for evaluating such methodologies is crucial before disaggregated data can be eventually used in fine-scale studies. In this paper, a new metric, named GDOWN, is proposed to assess the potential gain provided by disaggregation relative to the non-disaggregation case. The performance metric is tested during a four-year period by comparing 1-km resolution disaggregation based on physical and theoretical scale change (DISPATCH) data with the soil moisture measurements collected by six stations in central Morocco. DISPATCH data are obtained every 2–3 days from 40-km resolution SMOS (Soil Moisture Ocean Salinity) and 1-km resolution optical MODIS (Moderate Resolution Imaging Spectroradiometer) data. The correlation coefficient between GDOWN and the disaggregation gain in time series correlation, mean bias and bias in the slope of the linear fit ranges from 0.5 to 0.8. The new metric is found to be a good indicator of the overall performance of DISPATCH. Especially, the sign of GDOWN (positive in the case of effective disaggregation and negative in the opposite case) is independent of the uncertainties in SMOS data and of the representativeness of localized in situ measurements at the downscaling (1 km) resolution. In contrast, the traditional root mean square difference between disaggregation output and in situ measurements is poorly correlated (correlation coefficient of about 0.0) with the disaggregation gain in terms of both time series correlation and bias in the slope of the linear fit. The GDOWN approach is generic and thus could help test a range of downscaling methods dedicated to soil moisture and to other geophysical variables.

Common correlated effects estimation of heterogeneous dynamic panel data models with weakly exogenous regressors

This paper extends the Common Correlated Effects (CCE) approach developed by Pesaran (2006) to heterogeneous panel data models with lagged dependent variables and/or weakly exogenous regressors. We show that the CCE mean group estimator continues to be valid but the following two conditions must be satisfied to deal with the dynamics: a sufficient number of lags of cross section averages must be included in individual equations of the panel, and the number of cross section averages must be at least as large as the number of unobserved common factors. We establish consistency rates, derive the asymptotic distribution, suggest using covariates to deal with the effects of multiple unobserved common factors, and consider jackknife and recursive de-meaning bias correction procedures to mitigate the small sample time series bias. Theoretical findings are accompanied by extensive Monte Carlo experiments, which show that the proposed estimators perform well so long as the time series dimension of the panel is sufficiently large.

017 * INCREASED PERIOPERATIVE MORTALITY FOLLOWING APROTININ WITHDRAWAL: A REAL-WORLD ANALYSIS OF BLOOD MANAGEMENT STRATEGIES IN ADULT CARDIAC SURGERY

Objectives: Coagulopathic haemorrhage remains a major cause of perioperative morbidity in cardiac surgery. The serine protease inhibitor aprotinin prevents coagulopathic haemorrhage but was withdrawn in 2008 after clinical studies raised significant safety concerns. After careful re-analysis, aprotinin has been re-licensed for the prevention of coagulopathic haemorrhage. To better understand the risks and benefits of re-introducing aprotinin into practice, we aimed to determine the effect of the withdrawal of aprotinin on transfusion, bleeding, acute kidney injury (AKI) and mortality. Methods: Adult patients having cardiac surgery between 1 January 2005 and 31 December 2010 at a single tertiary cardiac centre, and for whom clinical, haematology and blood bank data were available, were included. Outcomes were compared between patients (n = 3587) operated on before aprotinin's withdrawal and propensity-matched patients (n = 3030) operated on after its withdrawal. Outcomes are presented as adjusted odds ratios (OR) or hazard ratios (HR) (95% confidence interval). Sensitivity analyses established the likely influence of regression, selection and time series bias on our results. Results: Aprotinin withdrawal was associated with significant increases in exposure to allogenic red blood cells: OR 1.24 (1.09-1.42), platelets 2.16 (1.81-2.58), fresh frozen plasma 1.81 (1.45-2.26) and cryoprecipitate 3.90 (1.69-8.98). Re-sternotomy for bleeding increased post-suspension, 2.10 (1.37-3.23), although blood loss on ITU was not significantly increased. Aprotinin withdrawal was associated with a significant increase in AKI incidence, 1.71 (1.51-1.93), and higher 30-day mortality rates, HR 2.10 (1.11-3.98). The results of sensitivity analyses did not alter our conclusions. Conclusions: The withdrawal of aprotinin resulted in increased allogenic blood component exposure, post-procedural AKI and mortality in cardiac surgery patients.

Common Correlated Effects Estimation of Heterogenous Dynamic Panel Data Models with Weakly Exogenous Regressors

This paper extends the Common Correlated Effects (CCE) approach developed by Pesaran (2006) to heterogeneous panel data models with lagged dependent variable and/or weakly exogenous regressors. We show that the CCE mean group estimator continues to be valid but the following two conditions must be satisfied to deal with the dynamics: a sufficient number of lags of cross section averages must be included in individual equations of the panel, and the number of cross section averages must be at least as large as the number of unobserved common factors. We establish consistency rates, derive the asymptotic distribution, suggest using covariates to deal with the effects of multiple unobserved common factors, and consider jackknife and recursive de-meaning bias correction procedures to mitigate the small sample time series bias. Theoretical findings are accompanied by extensive Monte Carlo experiments, which show that the proposed estimators perform well so long as the time series dimension of the panel is sufficiently large.

Increased perioperative mortality following aprotinin withdrawal: a real-world analysis of blood management strategies in adult cardiac surgery

To evaluate the effect of aprotinin withdrawal in 2008 on patient outcomes, to assess the likely risks and benefits of its re-introduction, and to consider the relevance of existing evidence from clinical trials to 'real-world' practice. We performed a nested case-control study of two cohorts undergoing adult cardiac surgery in a single tertiary centre. The first group underwent surgery between 1 January 2005 and 30 July 2007 (n = 3,578), prior to aprotinin withdrawal; the second group underwent surgery between 1 January 2009 and 31 December 2010 (n = 3,030), after aprotinin withdrawal. Propensity matching was used to select patients matched for 24 covariates in both groups (n = 3,508). We also estimated the effect of aprotinin withdrawal on a subgroup of high-risk patients (n = 1,002). Results were expressed as adjusted odds ratios (OR) and 95% confidence intervals (CI) for categorical data and hazard ratios (HR) for time-to-event data. In propensity-matched cohorts, the withdrawal of aprotinin from clinical use was associated with more bleeding, higher rates of emergency re-sternotomy, OR 2.10 (1.04-4.25), and acute kidney injury (AKI), OR 1.86 (1.53-2.25). In high-risk patients, the increases in bleeding and AKI following aprotinin withdrawal were of a greater magnitude. Aprotinin withdrawal was also associated with a significant increase in 30-day mortality, HR 2.51 (1.00-6.29), in the high-risk group. The results were not altered by sensitivity analyses that adjusted for potential selection bias, time series bias and unmeasured confounders. Aprotinin withdrawal was associated with increased complication rates and patient deaths following cardiac surgery. These real-world findings are at odds with those of randomised trials and cohort studies that have considered the clinical role of aprotinin.

Common Correlated Effects Estimation of Heterogeneous Dynamic Panel Data Models with Weakly Exogenous Regressors

This paper extends the Common Correlated Effects (CCE) approach developed by Pesaran (2006) to heterogeneous panel data models with lagged dependent variable and/or weakly exogenous regressors. We show that the CCE mean group estimator continues to be valid but the following two conditions must be satisfied to deal with the dynamics: a sufficient number of lags of cross section averages must be included in individual equations of the panel, and the number of cross section averages must be at least as large as the number of unobserved common factors. We establish consistency rates, derive the asymptotic distribution, suggest using co-variates to deal with the effects of multiple unobserved common factors, and consider jackknife and recursive de-meaning bias correction procedures to mitigate the small sample time series bias. Theoretical findings are accompanied by extensive Monte Carlo experiments, which show that the proposed estimators perform well so long as the time series dimension of the panel is sufficiently large.

Performance of a Bayesian state-space model of semelparous species for stock-recruitment data subject to measurement error

Measurement errors in spawner abundance create problems for fish stock assessment scientists. To deal with measurement error, we develop a Bayesian state-space model for stock-recruitment data that contain measurement error in spawner abundance, process error in recruitment, and time series bias. Through extensive simulations across numerous scenarios, we compare the statistical performance of the Bayesian state-space model with that of standard regression for a traditional stock-recruitment model that only considers process error. Performance varies depending on the information content in data, as determined by stock productivity, types of harvest situations, and amount of measurement error. Overall, in terms of estimating optimal spawner abundance S MSY , the Ricker density-dependence parameter β, and optimal harvest rate h MSY , the Bayesian state-space model works best for informative data from low and variable harvest rate situations for high-productivity salmon stocks. The traditional stock-recruitment model (TSR) may be used for estimating α and h MSY for low-productivity stocks from variable and high harvest rate situations. However, TSR can severely overestimate S MSY when spawner abundance is measured with large error in low and variable harvest rate situations. We also found that there is substantial merit in using h MSY (or benchmarks derived from it) instead of S MSY as a management target.

Cumulative effects of multiple sources of bias in estimating spawner–recruit parameters with application to harvested stocks of Chinook salmon ( Oncorhynchus tshawytscha)

Previous studies of bias in estimating spawner–recruit relationships have usually focused on a single source of bias, and have typically found bias that tends to overestimate productivity and underestimate optimal population size. However, many of these studies have used simulated data contrived to maximize the bias. This study examined the cumulative effects of multiple sources of bias using simulations with life history parameters based on Chinook salmon ( Oncorhynchus tshawytscha) stocks used as escapement indicator stocks for management of ocean salmon fisheries. Sources of bias incorporated in the simulations include time series bias, observational errors, and errors in estimating age composition. Cumulative biases in parameter estimates were found to be relatively small in most cases, and the resulting management reference points tended to err in the direction of conservation rather than in the direction of overexploitation. However, the same methods applied to depleted stocks would result in substantial bias that would tend to contribute to overexploitation.

Empirical and theoretical analyses of correction of time-series bias in stock-recruitment relationships of sockeye salmon (Oncorhynchus nerka)

Using data from 30 sockeye salmon (Oncorhynchus nerka) stocks and Monte Carlo simulations, we examined the importance of time-series bias on estimates of optimal harvest rate, optimal escapement, and sustainable yield. We compared the performance of the least-squares procedure for fitting a Ricker curve with an existing bias-correction method. Simulations showed that the effect of time-series bias is greatest for low-productivity stocks that exhibit a high degree of autocorrelation among residuals of the stock-recruitment relationship. A strong inverse empirical relationship between autocorrelation and stock productivity among the 30 stocks suggests that time-series bias is a more important concern for low-productivity northern stocks than for more productive southern stocks. The corrected method reduced bias in optimal escapement estimates under a limited set of conditions but at the price of increased variance in the estimates. For a constant escapement goal policy, using the bias correction thus resulted in sustainable yields slightly lower than or equal to expected values for 28 of the 30 stocks compared with yields obtained using the standard least-squares estimation method. We demonstrate the value of using a decision theoretic approach to evaluate the performance of estimation methods.

Time series bias in the estimation of density-dependent mortality in stock-recruitment models

Large biases can occur in parameter estimates for stock–recruitment models because the stock sizes are not chosen independently, being correlated with variability in recruitment. We examine the importance of this "time series bias" by a comprehensive analysis of available stock–recruitment data and the use of simulations. For semelparous species, i.e., species that reproduce only once, time series bias is important for all populations for which we had data. For iteroparous species, i.e., species that reproduce more than once, large biases occur if the populations are exploited at close to the maximum that is biologically possible. Notably, when there is autocorrelation in natural mortality, for univoltine species, the direction of bias is reversed due to model misspecification. Given moderate sample sizes and moderate levels of exploitation, time series bias is small for species such as Atlantic cod (Gadus morhua), for which α, the slope of the relationship between recruitment and number of spawners as the number of spawners goes to zero, is large. Time series bias will usually be important in species such as hakes (Merluccius) for which α appears to be relatively small.

Functional and Numerical Responses: A Framework for Fish‐Angler Interactions?

A conceptual framework for sport fisheries in which fish—angler interactions are viewed as a dynamic predator—prey relationship was evaluated with data from the walleye fishery in Lake Mendota, Wisconsin. A detailed 7—yr creel survey data set for this fishery revealed a link between fish and angler dynamics. Angler catch rates were positively and linearly related to fish abundance over the range of data in the study. Angling effort was positively related to catch rate. Fish—angler relationships were obscured by three difficulties commonly associated with fishery data sets: seasonal responses, measurement errors in predictor variables, and time series bias. Statistical techniques removed each of these biases, but the data were still insufficient to conclusively identify an appropriate model for the interaction. Nevertheless, it is clear that assumptions of static angler response to sport fish manipulations are unwarranted and capable of seriously confounding management efforts. Interactions among anglers and the influence of alternative fisheries outside the observed system appeared to be important. A combination of longer duration studies on a broader geographic scale and manipulative management experiments is probably necessary to advance our understanding of fish—angler interactions.

Factors Affecting the Time Series Bias in Stock–Recruitment Relationships and the Interaction Between Time Series and Measurement Error Bias

The stock–recruitment relationship (SRR) and the recruitment to spawning stock relationship (RSR) of the tiger prawn (Penaeus esculentus) in Exmouth Gulf, Western Australia, were used to examine the time series bias and measurement error bias by Monte Carlo simulations. The bias of the SRR parameters was decreased by adding environmental variables which reduced the variation about the SRR. However, because of the many variables which may be tested and the short time series of data usually available, one needs to be careful about possible spurious correlations. The variation about the RSR also needs to be taken into account, since it has an inverse relationship with the bias in the SRR. Simulating the effect of measurement error as well as the random variation about the SRR and RSR helps to determine the possible combinations of measurement error and random variation which may result in the observed variation about the SRR and RSR. These simulations also indicate that there is an interaction between measurement error and random variation about the SRR/RSR on the bias of the parameters of the SRR. This suggests that the time series bias and measurement error should be investigated simultaneously.

A Moving Cross-Section Analysis of Demand for Toothpaste

MRCA panel data on toothpaste expenditures are used to demonstrate how time series and cross-sectional bias can be eliminated by the method of covariance regression from single-equation demand least squares estimates.