Worst-case Bias Research Articles

On uniform confidence intervals for the tail index and the extreme quantile

This paper presents two results concerning uniform confidence intervals for the tail index and the extreme quantile. First, we show that there exists a lower bound of the length for confidence intervals that satisfy the correct uniform coverage over a nonparametric family of tail distributions. Second, in light of the impossibility result, we construct honest confidence intervals that are uniformly valid by incorporating the worst-case bias in the nonparametric family. The proposed method is applied to simulated data and real data of financial time series.

Lower bounds for the trade-off between bias and mean absolute deviation

In nonparametric statistics, rate-optimal estimators typically balance bias and stochastic error. The recent work on overparametrization raises the question whether rate-optimal estimators exist that do not obey this trade-off. In this work we consider pointwise estimation in the Gaussian white noise model with regression function f in a class of β-Hölder smooth functions. Let ’worst-case’ refer to the supremum over all functions f in the Hölder class. It is shown that any estimator with worst-case bias ≲n−β/(2β+1)≕ψn must necessarily also have a worst-case mean absolute deviation that is lower bounded by ≳ψn. To derive the result, we establish abstract inequalities relating the change of expectation for two probability measures to the mean absolute deviation.

Robust Mendelian randomization in the presence of residual population stratification, batch effects and horizontal pleiotropy

Mendelian Randomization (MR) studies are threatened by population stratification, batch effects, and horizontal pleiotropy. Although a variety of methods have been proposed to mitigate those problems, residual biases may still remain, leading to highly statistically significant false positives in large databases. Here we describe a suite of sensitivity analysis tools that enables investigators to quantify the robustness of their findings against such validity threats. Specifically, we propose the routine reporting of sensitivity statistics that reveal the minimal strength of violations necessary to explain away the MR results. We further provide intuitive displays of the robustness of the MR estimate to any degree of violation, and formal bounds on the worst-case bias caused by violations multiple times stronger than observed variables. We demonstrate how these tools can aid researchers in distinguishing robust from fragile findings by examining the effect of body mass index on diastolic blood pressure and Townsend deprivation index.

LowCon: A Design-based Subsampling Approach in a Misspecified Linear Model

We consider a measurement constrained supervised learning problem, that is, (i) full sample of the predictors are given; (ii) the response observations are unavailable and expensive to measure. Thus, it is ideal to select a subsample of predictor observations, measure the corresponding responses, and then fit the supervised learning model on the subsample of the predictors and responses. However, model fitting is a trial and error process, and a postulated model for the data could be misspecified. Our empirical studies demonstrate that most of the existing subsampling methods have unsatisfactory performances when the models are misspecified. In this paper, we develop a novel subsampling method, called “LowCon,” which outperforms the competing methods when the working linear model is misspecified. Our method uses orthogonal Latin hypercube designs to achieve a robust estimation. We show that the proposed design-based estimator approximately minimizes the so-called worst-case bias with respect to many possible misspecification terms. Both the simulated and real-data analyses demonstrate the proposed estimator is more robust than several subsample least-squares estimators obtained by state-of-the-art subsampling methods. Supplementary materials for this article are available online.

Impact of Varied Buffer Layer Designs on Single-Event Response of 1.2-kV SiC Power MOSFETs

In this article, the single-event response of the 1.2-kV silicon-carbide (SiC) power MOSFETs with varied buffer layer designs is investigated by the 2-D numerical simulations. The structural parameters of the buffer layers are compared and analyzed to understand the transient response after the heavy ion strike and the related physical mechanisms comprehensively. Simulation results reveal that an optimized single buffer structure can be acquired by using a relatively thicker buffer layer (T $\mu \text{m}$ ) and a moderate doping concentration (D cm−3). It demonstrates that the single-event-burnout (SEB) performance can be improved significantly under the worst case bias conditions [a drain bias voltage of 1.2 kV and a linear energy transfer (LET) of 1 pC/ $\mu \text{m}$ ]. In addition, the optimized structural combinations adopted in the dual or triple buffer layers can strengthen the SEB performance further. The simulation results show that a step electric field distribution is established inside the optimized dual or triple buffers, where the electric field peak is mitigated from 3 to 2.2 MV/cm. Moreover, the excess carrier generation is suppressed and the local temperature rise is weakened during the transient electrothermal process. Consequently, the structure with the optimal buffer layer designs can enlarge the SEB safe operating area (SOA) from 30%–50% to 100% rated breakdown voltage at high LET bias, which makes the SiC power MOSFETs used in aerospace and aviation power electronic systems become possible.

Accelerated Tests on Si and SiC Power Transistors with Thermal, Fastand Ultra-Fast Neutrons.

Neutron test campaigns on silicon (Si) and silicon carbide (SiC) power MOSFETs and IGBTs were conducted at the TRIGA (Training, Research, Isotopes, General Atomics) Mark II (Pavia, Italy) nuclear reactor and ChipIr-ISIS Neutron and Muon Source (Didcot, U.K.) facility. About 2000 power transistors made by STMicroelectronics were tested in all the experiments. Tests with thermal and fast neutrons (up to about 10 MeV) at the TRIGA Mark II reactor showed that single-event burnout (SEB) failures only occurred at voltages close to the rated drain-source voltage. Thermal neutrons did not induce SEB, nor degradation in the electrical parameters of the devices. SEB failures during testing at ChipIr with ultra-fast neutrons (1-800 MeV) were evaluated in terms of failure in time (FIT) versus derating voltage curves according to the JEP151 procedure of the Joint Electron Device Engineering Council (JEDEC). These curves, even if scaled with die size and avalanche voltage, were strongly linked to the technological processes of the devices, although a common trend was observed that highlighted commonalities among the failures of different types of MOSFETs. In both experiments, we observed only SEB failures without single-event gate rupture (SEGR) during the tests. None of the power devices that survived the neutron tests were degraded in their electrical performances. A study of the worst-case bias condition (gate and/or drain) during irradiation was performed.

Reduction of Bias from Parameter Variance in Geophysical Data Estimation: Method and Application to Ice Water Content and Sedimentation Flux Estimated from Lidar

Abstract This paper addresses issues of statistical misrepresentation of the a priori parameters (henceforth called ancillary parameters) used in geophysical data estimation. Parameterizations using ancillary data are frequently needed to derive geophysical data of interest from remote measurements. Empirical fits to the ancillary data that do not preserve the distribution of such data may induce substantial bias. A semianalytical averaging approach based on Taylor expansion is presented to improve estimated cirrus ice water content and sedimentation flux for a range of volume extinction coefficients retrieved from spaceborne lidar observations by CALIOP combined with the estimated distribution of ancillary data from in situ aircraft measurements of ice particle microphysical parameters and temperature. It is shown that, given an idealized distribution of input parameters, the approach performs well against Monte Carlo benchmark predictions. Using examples with idealized distributions at the mean temperature for the tropics at 15 km, it is estimated that the commonly neglected variance observed in in situ measurements of effective diameters may produce a worst-case estimation bias spanning up to a factor of 2. For ice sedimentation flux, a similar variance in particle size distributions and extinctions produces a worst-case estimation bias of a factor of 9. The value of the bias is found to be mostly set by the correlation coefficient between extinction and ice effective diameter, which in this test ranged between all possible values. Systematic reporting of variances and covariances in the ancillary data and between data and observed quantities would allow for more accurate observational estimates.

Observations on the bias of nonnegative mechanisms for differential privacy

We study two methods for differentially private analysis of bounded data and extend these to nonnegative queries. We first recall that for the Laplace mechanism, boundary inflated truncation (BIT) applied to nonnegative queries and truncation both lead to strictly positive bias. We then consider a generalization of BIT using translated ramp functions. We explicitly characterise the optimal function in this class for worst case bias. We show that applying any square-integrable post-processing function to a Laplace mechanism leads to a strictly positive maximal absolute bias. A corresponding result is also shown for a generalisation of truncation, which we refer to as restriction. We also briefly consider an alternative approach based on multiplicative mechanisms for positive data and show that, without additional restrictions, these mechanisms can lead to infinite bias.

Comparison of the Total Dose Responses of Fully Depleted SOI nMOSFETs With Different Geometries for the Worst Case Bias Conditions

Fully depleted (FD) silicon-on-insulator (SOI) nMOSFETs fabricated using 0.2- $\mu \text{m}$ SOI technology with external body contact (EBC) and floating body (FB) structures are irradiated up to 500 krad(Si) using 60Co gamma rays under the transmission gate (TG) and OFF bias conditions, respectively. The threshold voltage shift of the back transistors due to radiation-induced charge trapping in the buried oxide (BOX) is used to characterize the total dose response of irradiated back transistors. The results indicate that, overall, the TG biased EBC transistors with various gate lengths are more sensitive to the total dose radiation than the OFF biased FB counterparts at low dose levels. Furthermore, for the shorter gate transistors with a supply voltage of 1.8 V, a relatively complex correlation exists between the total dose response and the gate length. In particular, the back transistor of the TG biased 0.20- $\mu \text{m}$ 1.8-V EBC transistor reveals smaller threshold voltage shifts than the OFF biased counterpart at approximately 500 krad(Si) because of the saturation of its threshold shift with increased doses, and a similar radiation response is observed for the TG biased 0.35- $\mu \text{m}$ EBC transistor with a supply voltage of 2.5 V. We found that the 2.5-V back transistors exhibited greater radiation hardness than the 1.8-V counterparts with the same bias configuration in most cases.

Charge Buildup and Spatial Distribution of Interface Traps in 65-nm pMOSFETs Irradiated to Ultrahigh Doses

In this paper, a commercial 65-nm CMOS technology is irradiated at ultrahigh ionizing doses and then annealed at high temperature under different bias conditions. The experimental results demonstrate the high sensitivity of pMOSFETs to radiation-induced short-channel effects, related to the buildup of defects in spacer dielectrics. We find that the charge buildup in the spacers is insensitive to the applied source-to-drain electric field, but the generation and/or annealing of interface traps strongly depends on applied drain bias and channel length. The static dc and charge-pumping measurements indicate a high density of interface traps in the lateral source/drain extension regions. The worst case bias condition corresponds to the application of a large drain–source voltage, due to the lateral electric field driving hydrogen from the spacers toward the source extension and the channel. The consequent differences in growth and annealing rates of interface traps lead to a large asymmetric degradation of the short-channel transistors. The technology computer-aided design simulations are used to qualitatively confirm the proposed degradation model.

On the Informativeness of Descriptive Statistics for Structural Estimates

We propose a way to formalize the relationship between descriptive analysis and structural estimation. A researcher reports an estimate ĉ of a structural quantity of interest c that is exactly or asymptotically unbiased under some base model. The researcher also reports descriptive statistics γ that estimate features γ of the distribution of the data that are related to c under the base model. A reader entertains a less restrictive model that is local to the base model, under which the estimate ĉ may be biased. We study the reduction in worst-case bias from a restriction that requires the reader's model to respect the relationship between c and γ specified by the base model. Our main result shows that the proportional reduction in worst-case bias depends only on a quantity we call the informativeness of γ for ĉ. Informativeness can be easily estimated even for complex models. We recommend that researchers report estimated informativeness alongside their descriptive analyses, and we illustrate with applications to three recent papers.

Analysis and Mitigation of Bias Injection Attacks Against a Kalman Filter

In this paper, we consider a state estimation problem for stochastic linear dynamical systems in the presence of bias injection attacks. A Kalman filter is used as an estimator, and a chi-squared test is used to detect anomalies. We first show that the impact of the worst-case bias injection attack in a stochastic setting can be analyzed by a deterministic quadratically constrained quadratic program, which has an analytical solution. Based on this result, we propose a criterion for selecting sensors to secure in order to mitigate the attack impact. Furthermore, we derive a condition on the necessary number of sensors to secure in order for the impact to be less than a desired threshold.

Sensitivity analysis for matched pair analysis of binary data: From worst case to average case analysis.

In matched observational studies where treatment assignment is not randomized, sensitivity analysis helps investigators determine how sensitive their estimated treatment effect is to some unmeasured confounder. The standard approach calibrates the sensitivity analysis according to the worst case bias in a pair. This approach will result in a conservative sensitivity analysis if the worst case bias does not hold in every pair. In this paper, we show that for binary data, the standard approach can be calibrated in terms of the average bias in a pair rather than worst case bias. When the worst case bias and average bias differ, the average bias interpretation results in a less conservative sensitivity analysis and more power. In many studies, the average case calibration may also carry a more natural interpretation than the worst case calibration and may also allow researchers to incorporate additional data to establish an empirical basis with which to calibrate a sensitivity analysis. We illustrate this with a study of the effects of cellphone use on the incidence of automobile accidents. Finally, we extend the average case calibration to the sensitivity analysis of confidence intervals for attributable effects.

Worst-Case Bias for Proton and 10-keV X-Ray Irradiation of AlGaN/GaN HEMTs

Responses to 1.8 MeV proton irradiation and 10-keV X-ray irradiation under typical bias conditions are investigated for AlGaN/GaN HEMTs fabricated with different types of process technologies. We find that, in contrast to previous generations of process technologies, total ionizing dose effects can be significant in these devices. For proton irradiation, worst-case bias for transconductance degradation for GaN-on-SiC substrate devices is ON bias, and for devices built on free-standing GaN substrates, the worst-case bias condition is semi-ON bias. Low-frequency noise measurements demonstrate that these differences result from differences in defect types and energy distributions for the different types of devices, both before and after irradiation. These results emphasize the need to test devices under a wide range of conditions during characterization and qualification testing.

Dose-rate sensitivity of deep sub-micro complementary metal oxide semiconductor process

Enhancing low dose rate sensitivity (ELDRS) in bipolar device is a major problem of liner circuit radiation hardness prediction for space application. ELDRS is usually attributed to space-charge effect. A key element is the difference in transport rate between holes and protons in SiO2. Interface-trap formation at high dose rate is reduced due to positive charge buildup in the Si/SiO2 interfacial region (due to the trapping of holes and/or protons) which reduces the flow rates of subsequent holes and protons (relative to the low-dose-rate case) from the bulk of the oxide to the Si/SiO2 interface. Generally speaking, the dose rate of metal oxide semiconductor (MOS) device is time dependent when annealing of radiation-induced charge is taken into account. The degradation of MOS device induced by the low dose rate irradiation is the same as that by high dose rate when annealing of radiation-induced charge is taken into account. However, radiation response of new generation MOS device is dominated by charge buildup in shallow trench isolation (STI) rather than gate oxide as older generation device. Unlike gate oxides, which are routinely grown by thermal oxidation, field oxides are produced using a wide variety of deposition techniques. As a result, they are typically thick (100 nm), soft to ionizing radiation, and electric field is far less than that of gate oxide, which is similar to the passivation layer of bipolar device and may lead to ELDRS. Therefore, dose-rate sensitivities of n-type metal oxide semiconductor field effect transistor (NMOSFET) and static random access memory (SRAM) manufactured by 0.18 m complementary metal oxide semiconductor (CMOS) process are explored experimentally and theoretically in this paper. Radiation-induced leakages in NMOSFET and SRAM are examined each as a function of dose rate. Under the worst-case bias, the degradation of NMOSFET is more severe under the low dose rate irradiation than under the high dose rate irradiation and anneal. Moreover, radiation-induced standby current rising in SRAM is more severe under the low dose rate irradiation than under the high dose rate irradiation even when anneal is not considered. The above experimental results reveal that the dose-rate sensitivity of deep sub-micron CMOS process is not related to time-dependent effects of CMOS devices. Mathematical description of the combination between enhanced low dose-rate sensitivity and timedependent effects as applied to radiation-induced leakage in NMOSFET is developed. It has been numerically found that non time-dependent effect of deep sub-micron CMOS device arises due to the competition between enhanced low dose-rate sensitivity in bottom of STI and time-dependent effect at the top of STI. The high dose rate irradiation is overly conservative for devices used in a low dose rate environment. The test method provides an extended room temperature anneal test to allow leakage-related parameters that exceed postirradiation specifications to return to a specified range.

Bias dependence of TID induced single transistor latch for 0.13 μm partially depleted SOI input/output NMOSFETs

In this work, single transistor latch effects induced by total dose irradiation for 0.13μm partially depleted silicon-on-insulator (PDSOI) n-type metal-oxide-semiconductor field effect transistors (NMOSFETs) were investigated. The front gate transfer characteristics under different bias configurations with forward and reverse gate voltage sweep are characterized to evaluate the latch phenomenon. The results indicate that transmission–gate (TG) bias is the worst case bias for total dose induced latch, and the onset drain voltage required for latchup degrades as the irradiation level increased. Experiments and 2D simulations are performed to analyze the positive trapped charge in the buried oxide (BOX) and its impact on the latch effect. It is demonstrated that the irradiation can enhance the impact ionization and thereby make the device more sensitive to latchup, especially at negative gate voltage. Moreover, the radiation induced coupling effect between the front gate and back gate can make the PDSOI devices in our experiments behave like the fully depleted (FD) ones.

Impact of bias conditions on electrical stress and ionizing radiation effects in Si-based TFETs

The interplay between electrical stress and ionizing radiation effects is experimentally investigated in Si-based Tunnel Field Effect Transistors (TFETs). In particular, the impact of bias conditions on the performance degradation is discussed. We found that the electrical stress effects in TFETs could not be ignored in radiation tests, since they can possibly overwhelm the radiation-induced degradation. Under this circumstance, the worst-case bias condition for studying radiation effects is not straightforward to be determined when there is an interplay between electrical stress and ionizing radiation effects.

A Simulation Study of Hot Carrier Effects in SoI-Like Bulk Silicon nMOS Device

A silicon-on-insulator (SoI)-like bulk silicon (SLBS) MOSFET structure is studied and compared against a fully depleted (FD) SoI MOSFET using 2-D numerical simulations. A p/n - /p + structure was contained in SLBS device, in which n - is made of 4H-SiC. This n - layer is FD. Hot carrier (HC) effects (HCEs) in proposed SLBS nMOSFET were simulated and compared with that in FD SoI nMOSFET. In this paper, HC-induced device degradation is characterized under different temperatures. HCE of SLBS always has an advantage over that of SoI. The worst case bias condition for HCEs has also been studied and was found as V gs = 1/2V ds .

A New Approach to Calculate the Vertical Protection Level in A-RAIM

Four methods to calculate the Vertical Protection Level (VPL) can be used in Advanced Receiver Autonomous Integrity Monitoring (A-RAIM), among which the ideal method is the strictest one. To obtain the ideal VPL satisfying the exact required integrity risk, the worst case bias with the maximum integrity risk is searched for. This investigation has found that the correct worst case highly depends on the choice of the input VPL. To gain the correct result, the computation becomes complex and the accuracy of the result is compromised. Therefore, a new procedure is designed with a new search: the maximum VPL is searched to encompass all possible bias sizes. Since VPL is calculated with a given integrity risk for each bias size, the uncertainty of the arbitrary VPL input in the ideal method is avoided. Also, an optimisation algorithm is adopted to improve computational efficiency. It is shown that the new method is more reliable and efficient than the current best method. Simulation results worldwide also show that the new approach has improved A-RAIM availability from 32%–38% to 74% with GPS and from 44%–43% to 85% with Galileo.

The total ionizing dose effect in 12-bit, 125 MSPS analog-to-digital converters

This paper presents the total ionizing dose test results at different biases and dose rates for AD9233, which is fabricated using a modern CMOS process. The experimental results show that the digital parts are more sensitive than the other parts. Power down is the worst-case bias, and this phenomenon is first found in the total ionizing dose effect of analog-to-digital converters. We also find that the AC as well as DC parameters are sensitive to the total ionizing dose at a high dose rate, whereas none of the parameters are sensitive at a low dose rate. The test facilities, results and analysis are presented in detail.