0

I am conducting meta-analysis and applied funnel plot to assess for publication bias using metafor. First, I tried to explore the dataset to identify the outliers by creating funnel plots with 1/SE using the following code

par(mfrow = c(1, 2))
funnel(Soil_total_NPK$lnrr, Soil_total_NPK$v, yaxis="seinv",
       #xlim = c(-3, 3),
       ylab = "Precision (1/SE)",
       xlab = "Effect size (RR)", col =Soil_total_NPK$unique_id, atransf = exp)

It seems there are some outliers due to the very small values in the variance of effect size. Does anyone know how to create a criterion to remove very minimum values in the variance? I tried to use the following code but it only removed the only one smallest value.

Soil_total_NPK1 <- Soil_total_NPK[Soil_total_NPK$v != min(Soil_total_NPK$v), ]
madina_b
  • 35
  • 5
  • 1
    How exactly are you defining "very minimum values in the variance"? It's easier to help you if you include a simple [reproducible example](https://stackoverflow.com/questions/5963269/how-to-make-a-great-r-reproducible-example) with sample input and desired output that can be used to test and verify possible solutions. If you are looking for data analysis advice rather than programming advice you should ask somewhere like [stats.se] rather than Stack Overflow. – MrFlick Dec 12 '22 at 16:35
  • 1
    I don't know if what you target is 'outliers', also if you remove 'outliers' (and I write it with quotes for a good reason), your funnel plot will be updated and you could possibly get new 'outliers' – Yacine Hajji Dec 12 '22 at 16:42

0 Answers0