Question

我试图多次重新采样（替换）一个data.frame。该数据框是按场地矩阵的物种。此数据中有几个零，这意味着重采样过程有时只产生零行。因此，我想从每个for循环中消除只有零行的矩阵，因为函数multipart（vegan package）不接受这些矩阵。

please, download data here

提前谢谢大家。

require(vegan)

# DATA (Edited)

dat <- read.table("df.stack.txt", h=T) # Link above
mat <- dat[,c(-1, -2, -3)]
levels <- dat[,1:3]

## Vectors for saving desired statistics
alfa.null <- NULL
beta1.null <- NULL
beta2.null <- NULL

## Shuffling matrices ## Here is the problem

times <- 99
for (i in times) {
mat.rand <- sample(mat, replace=TRUE) ### Problem to be solved - some matrices will have only zero rows 
div.aleat <- multipart(mat.rand ~ ., levels, index="renyi", scales=1, nsimul=99, global=TRUE)
alfa.null[i] <- div.aleat$statistic[[1]]
beta1.null[i] <- div.aleat$statistic[[4]]
beta2.null[i] <- div.aleat$statistic[[5]]
}

Answer 1

你需要保持重新采样，直到得到如此的非零矩阵：

times <- 99
for (i in seq.int(times)) {
    while(TRUE){
        mat.rand <- sample(df, replace=TRUE) ### Problem to be solved - some matrices will have only zero rows 
        if(any(rowSums(mat.rand)!=0))
            break
    }

    blah blah blah
}

当然使用while(TRUE)循环，抓住无限循环是个好主意：

times <- 99
max_Tries <- 100
for (i in seq.int(times)) {
    tries = 0
    while(TRUE){
        mat.rand <- sample(df, replace=TRUE) ### Problem to be solved - some matrices will have only zero rows 
        if(any(rowSums(mat.rand)!=0))
            break
        tries <- tries + 1
        if(tries > max_Tries)
            stop("number of tries exceeded max_Tries: possible infinite loop")
    }

    blah blah blah
}

在洗牌data.frame之后仅删除零行

1 个答案: