Shulou(Shulou.com)06/01 Report--

How to determine the best logistic regression equation, I believe that many inexperienced people do not know what to do. Therefore, this paper summarizes the causes and solutions of the problem. Through this article, I hope you can solve this problem.

Question: Hello, I would like to ask how to choose the test method in the multivariate analysis of binary logistic regression. Do you choose the most commonly used "backward (conditional)", or try each test method and then choose a test method according to what indicators?

Regression model fitting, we can regard regression fitting as a kind of "exploratory" analysis, that is, to explore the optimal regression equation through regression fitting, so can logistic regression. Therefore, according to our professional knowledge, we can determine which variables need to be introduced into the logistic regression equation, of course, we can also use different stepwise regression methods to fit, judge the advantages and disadvantages of the regression equation according to various statistics, and find out the "optimal" regression equation. To sum up, there are roughly the following categories of indicators

1. Goodness-of-fit test of the model

If the predicted value is close to the actual observed value, it shows that the fitting effect of the model is good, the value of statistics is small, and the corresponding P value is larger. Test hypothesis H0: the fitting effect of the model is good, and α can be 0.1 or 0.2.

1. Deviation test, the statistic is χ D2.

2. Pearson χ 2 test, the statistic is χ 2

3. Homser-Lemeshow test, χ HL2

Judgment principle: the smaller the χ 2 value is, the larger the P is, the better the fitting effect of the model is.

2. Information index of model goodness-of-fit

1.-2lnL.

2. AIC criterion.

3. SC criterion.

When other conditions remain the same, the smaller the three indicators are, the better the model fits.

3. Prediction accuracy of logistic regression model.

1. Generalized determination coefficient: Cox-Snell generalized determination coefficient (Cox-SnellR2) and Nagelkerke generalized determination coefficient (Nagelkerke R2) are similar to the determination coefficient R2 in linear regression analysis. The values of these two indexes are between 0 and 1. The larger the index, the greater the proportion of the variation explained by the model, the higher the accuracy of the model prediction.

2. The correlation between prediction probability and observation, the commonly used evaluation indicators are Somers'D, Goodman-Kruskal Gamma,

Kendall's Tau-an and KendaIl'sTau-c, the larger the absolute value of the index, the higher the correlation between the prediction probability and the response variables, which means that the prediction ability of the model is stronger.

4. Prediction accuracy

Used to evaluate the prediction accuracy of logistic regression model, the higher the accuracy, the better the effect.

Fifth, the choice of the "best" model

Readers can make a comprehensive evaluation of the fitted model combined with the above statistical indicators, and finally select the "best" model from the aspects of interpretability, simplicity and accessibility of variables on the basis of statistical analysis and professional knowledge. Usually, the "best" model can not be determined by one calculation, but is determined by different combination analysis of variables.

After reading the above, have you mastered how to determine the best logistic regression equation? If you want to learn more skills or want to know more about it, you are welcome to follow the industry information channel, thank you for reading!

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.