英文:
modelsummary display output from lm and brms concurrently with statistics
问题
models <- list("Linear" = lm(outcome ~ week * food data = df ),"Bayesian" = brm(outcome ~ s(week, k = 4, fx = TRUE, by = food) + food, data = df, family = "zero_one_inflated_beta"))
以下代码在运行时有效：
modelsummary(models,estimate = "{estimate}[{conf.low}, {conf.high}]",statistic = NULL)
问题在于，当我尝试使用以下代码获取线性模型的p值、t值和标准误差时，出现错误，错误信息为“Error: std.error is not available. The estimate and statistic arguments must correspond to column names in the output of this command: get_estimates(model)”
modelsummary(models,estimate = "estimate[{conf.low}, {conf.high}]",statistic = c("Std.Error" = "std.error","t-value" = "statistic","p-value" = "p.value"))
如何使modelsummary()在没有统计信息的情况下（例如在brms模型中）忽略统计信息的显示，而不是抛出错误呢？
英文:
models <- list("Linear" = lm(outcome ~ week * food data = df ),"Bayesian" = brm(outcome ~ s(week, k = 4, fx = TRUE, by = food) + food, data = df, family = "zero_one_inflated_beta"))
The following code works when I run it
modelsummary(models,estimate = "{estimate}[{conf.low}, {conf.high}]",statistic = NULL)
The problem is that when I attempt to also get the p-value, t-value and standard error of the linear model with the following code, the error comes up as Error: std.error is not available. The estimate and statistic arguments must correspond to column names in the output of this command: get_estimates(model)
modelsummary(models,estimate = "estimate[{conf.low}, {conf.high}]",statistic = c("Std.Error" = "std.error","t-value" = "statistic","p-value" = "p.value"))
How can I enable modelsummary() to ignore the display of statistics when there is none like in the case of the brms model instead of throwing an error?
答案1
得分: 2
这是一个常见的用例，在CRAN版本的modelsummary中不受很好支持。而不是建议一个复杂的解决方法，我推送了一个更简单的解决方法到开发版本中。您可以使用以下方法立即安装它：
remotes::install_github("vincentarelbundock/modelsummary")
为使更改生效，完全重启R。
然后，您可以执行类似以下的操作：
library(brms)library(modelsummary)mod1 <- lm(mpg ~ hp + qsec, data = mtcars)mod2 <- brm(mpg ~ hp + qsec, data = mtcars)models <- list(mod1, mod2)modelsummary(models,statistic = c("std.error", "conf.int"),# 清除系数名称coef_rename = \(x) gsub("b_", "", x),coef_omit = "Intercept")
| (1) | (2) | |
|---|---|---|
| hp | -0.085 | -0.084 |
| (0.014) | ||
| [-0.113, -0.056] | [-0.112, -0.055] | |
| qsec | -0.887 | -0.867 |
| (0.535) | ||
| [-1.980, 0.207] | [-1.944, 0.239] | |
| sigma | 3.815 | |
| [3.000, 4.991] | ||
| Num.Obs. | 32 | 32 |
| R2 | 0.637 | 0.631 |
| R2 Adj. | 0.612 | 0.553 |
| AIC | 180.3 | |
| BIC | 186.2 | |
| Log.Lik. | -86.170 | |
| F | 25.431 | |
| ELPD | -91.1 | |
| ELPD s.e. | 4.9 | |
| LOOIC | 182.3 | |
| LOOIC s.e. | 9.8 | |
| WAIC | 181.9 | |
| RMSE | 3.57 | 3.57 |
英文:
This is a common use-case which is not well supported in the CRAN version of modelsummary. Instead of suggesting a complicated hack, I pushed a change to the development version which makes this much easier. You can install it now with:
remotes::install_github("vincentarelbundock/modelsummary")
Restart R completely for the changes to take effect.
Then, you can do things like:
library(brms)library(modelsummary)mod1 <- lm(mpg ~ hp + qsec, data = mtcars)mod2 <- brm(mpg ~ hp + qsec, data = mtcars)models <- list(mod1, mod2)modelsummary(models,statistic = c("std.error", "conf.int"),# clean-up coefficient namescoef_rename = \(x) gsub("b_", "", x),coef_omit = "Intercept")
| (1) | (2) | |
|---|---|---|
| hp | -0.085 | -0.084 |
| (0.014) | ||
| [-0.113, -0.056] | [-0.112, -0.055] | |
| qsec | -0.887 | -0.867 |
| (0.535) | ||
| [-1.980, 0.207] | [-1.944, 0.239] | |
| sigma | 3.815 | |
| [3.000, 4.991] | ||
| Num.Obs. | 32 | 32 |
| R2 | 0.637 | 0.631 |
| R2 Adj. | 0.612 | 0.553 |
| AIC | 180.3 | |
| BIC | 186.2 | |
| Log.Lik. | -86.170 | |
| F | 25.431 | |
| ELPD | -91.1 | |
| ELPD s.e. | 4.9 | |
| LOOIC | 182.3 | |
| LOOIC s.e. | 9.8 | |
| WAIC | 181.9 | |
| RMSE | 3.57 | 3.57 |
通过集体智慧和协作来改善编程学习和解决问题的方式。致力于成为全球开发者共同参与的知识库，让每个人都能够通过互相帮助和分享经验来进步。
评论