Goodman N W, Curnow J S
Br J Anaesth. 1985 Mar;57(3):311-8. doi: 10.1093/bja/57.3.311.
The ventilatory response to carbon dioxide using the rebreathing method was studied repeatedly in four healthy volunteers. The data were collected and processed in different ways using a microcomputer. Least squares linear regression analysis described the responses almost as accurately as a non-linear second-order polynomial regression. Linear regression analysis was undertaken on VI/PE'CO2 points, first breath-by-breath and then on three different groupings of breaths. The gradients and intercepts obtained by grouping agreed with those from breath-by-breath analysis both for a single response and for the mean result of a number of responses. Grouping tended to increase the correlation coefficient of a response, but widened the statistical confidence limits of the calculated gradient. The performance variability between repeated responses, the same no matter which data points were used, produced a range of gradients for each subject which was larger than the confidence limits on the gradient of a single response.
我们对四名健康志愿者反复进行了研究，采用再呼吸法测定其对二氧化碳的通气反应。使用微型计算机以不同方式收集和处理数据。最小二乘线性回归分析对反应的描述几乎与非线性二阶多项式回归一样准确。对VI/PE'CO2点进行线性回归分析，首先逐次呼吸进行，然后对三组不同的呼吸进行分析。分组得到的梯度和截距与逐次呼吸分析得到的结果一致，无论是单个反应还是多个反应的平均结果。分组往往会增加反应的相关系数，但会扩大计算梯度的统计置信区间。无论使用哪些数据点，重复反应之间的性能变异性都是相同的，这使得每个受试者的梯度范围大于单个反应梯度的置信区间。
