Search

Probability predictions from binary regressions or machine learning methods ought to be calibrated: if an event is predicted to occur with probability |$x$|⁠ , it should materialize with approximately that frequency, which means that the so-called calibration curve |$p(\cdot)$| should equal the identity, i.e. |$p(x) = x$| for all |$x$| in the unit interval. We propose honest calibration assessment based on novel confidence bands for the calibration curve, which are valid subject to only the natural assumption of isotonicity. Besides testing the classical goodness-of-fit null hypothesis of perfect calibration, our bands facilitate inverted goodness-of-fit tests whose rejection allows for the sought-after conclusion of a sufficiently well-specified model. We show that our bands have a finite-sample coverage guarantee, are narrower than those of existing approaches, and adapt to the local smoothness of the calibration curve |$p$| and the local variance of the binary observations. In an application to modelling predictions of an infant having low birth weight, the bounds give informative insights into model calibration. [ABSTRACT FROM AUTHOR]

A system‐level method to validate previous calibration results for low‐precision inertial measurement unit (IMU) in personal navigation system (PNS) is proposed. Micro‐electro mechanical systems inertial sensors have gained popularity in PNS because they are small, compact and inexpensive. However, the calibration results of these sensors which belong to low‐grade IMU will change as time and environment change. It is significant to know whether previous calibration results are reliable for navigation after a long time. However, users have no devices such as turntable to validate calibration results precisely. Therefore, a triple two‐position method without any external aiding is presented to validate the bias errors relative to previous calibration results. If the errors are less than the bias range which is a priori information calculated in previous lab test, previous calibration results are available, otherwise the calibration results is invalid. The two‐position method based on the strong tracking filter, which takes velocity and azimuth errors as observations, can estimate the bias and scale factor errors of accelerometers and gyros. The experimental results show that the proposed method is effective, compared with the conventional 24‐position calibration method on the turntable and the system‐level multi‐position method usually used for high‐grade IMU. [ABSTRACT FROM AUTHOR]

InfraRed Atmospheric Sounder (IRAS) instruments were successfully launched onboard the FengYun-3A (FY3A) and FengYun-3B (FY3B) satellites on May 27, 2008, and November 5, 2010, respectively. They aim at providing multichannel radiances within the spectral range of visible to infrared (IR) wavelengths for many environmental applications, including data assimilation and retrievals of global atmospheric temperature and humidity profiles. However, the velocity of the filter wheel of the first IRAS onboard FY3A is unstable and, therefore, induced a discontinuity in the measurement. The IRAS onboard FY3B works well in normal and stable operational mode since its launch without any anomaly. A variety of postlaunch calibration/validation tasks are conducted using on-orbit data during a period of three months. This paper presents on-orbit verification of IRAS instrument performance, including long-term trends of the space and warm calibration counts and noise equivalent delta radiance. The Earth scenes observed simultaneously by IRAS and Meteorological Operational Satellite Programme (METOP)/Infrared Atmospheric Sounding Interferometer were obtained and compared to demonstrate a close similarity between the two measurements. Furthermore, the IR channel observations from FY3B/IRAS are compared with those from National Oceanic and Atmospheric Administration-19/High Resolution Infrared Radiation Sounder (HIRS) equivalent channels and simulations from a radiative transfer model. The results show that some of IRAS IR channels perform very well, particularly for channels 1-10, 15, 19, and 20, compared to those of HIRS. Several channels, such as 13, 16, and 18, however, display some large biases. The causes of these increased biases are still under investigation. [ABSTRACT FROM AUTHOR]