有些文献里，使用弹性网络回归在测试集构建预测模型，在内外部验证集验证时对截距进行了重新校准，这是如何办到的【文献中并没有详细说明】？Therefore, after applying the age predictor to the testing datasets, we fitted a linear regression of methylation age = chronological age + population intercept in theWHZH, COW, and SY panels, separately; we then recalibrated the model intercept of SY and COW to make the population intercept of SY and COW the same as the population intercept of WHZH (referred to as intercept equalization in the following text).

The WHZH, SY, COW, and TwinS were process in the same lab; therefore, intercepts were adjusted using the intercept equalization method; the SLSJ a separate lab and 5 CpGs in the age predictor were not available due to failure of quality control; therefore, the intercept were adjusted using a least RMSE method.