Statistics bias prevention

Refer to Section 6.4.7 and Exhibit 6.12 to explain randomization and the techniques used. Randomization and double-blinding are necessary to prevent bias in data collection so that statistical analysis based on normal distribution can be used to evaluate the trial results. 

The optimal model is determined by finding the minimum error between the extracted concentrations and the reference concentrations. Cross-validation is also used to determine the optimal number of model parameters, for example, the number of factors in PLS or principal components in PCR, and to prevent over- or underfitting. Technically, because the data sets used for calibration and validation are independent for each iteration, the validation is performed without bias. When a statistically sufficient number of spectra are used for calibration and validation, the chosen model and its outcome, the b vector, should be representative of the data. 

The tests must be statistically designed to prevent bias. 

Cross-sectional studies can identify prevalence rates based on the distribution of a particular syndrome or malformation, but the study design makes it difficult to identify cause and effect relationships. Case-control protocols match the affected pregnancy to an unaffected pregnancy but, again, it is difficult to account for maternal recall bias and perhaps equally difficult to control for bias (even unconscious) on the part of the investigator. Cohort smdies, while suffering from problems of dose determination, are usually prospective, the largest, the most expensive, the slowest, and usually the most statistically powerful to detect reliable associations. Double-blind intervention studies, like those conducted with folic acid and prevention of NTD, usually yield the most conclusive data. 

The bathtub curve shows three areas over time. The early failure phase describes the time frame in which the failure behavior is not sufficiently developed through unknown influences, environment parameters, correct materials and bias points. This should be investigated for the development of components within the context of the design verification so that phase 2, the usage phase, can be entered at the beginning of series production. The usage phase should be designed in a way that the failure rate only starts after the expiration of the statistical life expectation of the components. In reality the failure rate is placed below the bathtub curve, as far as necessary so that an age induced increase can be seen and a sufficient robustness level ensures that the statistical life expectation is achieved. ISO 26262 does not mention any requirements, for example in order to prevent early failure behavior. 

---