《Table 3 Comparison of the results by different methods for the three datasets》

  提示：宽带有限、当前游客访问压缩模式

本系列图表出处文件名：随高清版一同展现

《A variable importance criterion for variable selection in near-infrared spectral analysis》

1. 获取 高清版本忘记账户？点击这里登录

1. 下载图表忘记账户？点击这里登录

a) Number of spectral variables；b) number of latent variables；c) mean value and the standard deviation over 500 independent runs of the robustness test.

To verify the rationality of the C value，five spectral bands with higher（than Q3）and lower（than Q1）C value are investigated.As indicated by the red and blue bars in the figure，the spectral intervals of 1228–1242，1292–1310，1470–1486，1280–1288 and 1490–1550 nm were studied.The former three bands should have positive contributions to the model，and the latter two bands should have negative effects on the model.Taking the full spectra model as a reference，the results obtained by removing the variables in the five spectral intervals are compared in Table 1.