《学术论文集锦[开错]》紫微客

7、007

　　Secondly, to obtain the range of cell inhibition rates corresponding to a specific dose, we need to estimate confidence intervals for the sample data. In general, we set the confidence level 1- as 95%, i.e. the significance level . In this study, I used this method to estimate the 95% confidence interval for the mean cell inhibition rate corresponding to a specific dose. Each patient has a different response to the drug, so I took the range as an indication of the overall response of the patient population.

　　In addition, as a classical dose-response model, Four-parameter Logistic Regression model is also one of the key research objects in this study, and its mathematical expression is shown as follows:

　　(1)

　　Where,
　　c represents the lowest point of the curve, and its practical significance is the degree of response without drug, E0.
　　d represents the highest point of the curve, and its practical significance is the maximum response of cell to a drug, Emax.
　　b represents a normalized value of slope and is not the actual slope of the curve.
　　e represents the concentration of a drug at which 50% of its maximal response, EC50.
　　Then, is the 50% of its maximal response is equal to 50%, e also is the IC50 of the drug. Otherwise, the IC50 can be computed substituting f(x) for 50 and finding x in the curve equation:
　　(2)

　　This model is an ideal dose-response model, but in the actual fitting process, I found that some data of the experimental group could not be analyzed by this model. We must ensure the independence, normality and variance constancy of the data. Only when these three premises are satisfied, the results of the calculation can be reliable. However, it is obvious that not all the sample data used in this study can meet the above conditions.

　　In short, the above classical statistical analysis methods are more suitable for large samples, while the data set of this study is small sample, which means that the results of statistical analysis are not necessarily reliable (such as normality test, ANOVA and confidence interval, etc.). At the same time, in order to find suitable research approach of dose-response, we must introduce some methods suitable for small sample data analysis to improve these classical statistical theories, which will be introduced detailed in Part 4.3.