Using a function I made to conduct six tests from 5000 datasets of N(0,1)-CodePudding

I'm trying to use this function I made

ev <- function(x, mu0 = y, alt = "two-sided", lev = 0.95) {
  ts<- ((mean(x)-mu0)/(sd(x)/sqrt(length(x))))
  if(alt == "two-sided") {
    pval <- 2*pt(ts, length(x-1))
  ci<- c((mu0- pval*(length(x)/ts)), mu0   pval*(length(x)/ts))
  }
  else if(alt == "greater"){
    pval <- (1- pt(ts, length(x-1)))
    ci<- c((mu0- pval*(length(x)/ts)), mu0   pval*(length(x)/ts))
  }
  else if(alt == "less") {
    pval <- pt(ts, length(x-1))
    ci<- c((mu0- pval*(length(x)/ts)), mu0   pval*(length(x)/ts))
  }
    return(list(c(ts,pval, ci)))
}

to solve the following problem.

I created the datasets using

g<- matrix(rnorm(5000*20),nrow = 20, ncol = 5000)
my.list<- split(g, rep(1:ncol(g), each= nrow(g)))

But am a little confused about what to do from here. I tried just plugging in my.list, but that gives me this error

ev(my.list,0.2,"less")
Error in var(if (is.vector(x) || is.factor(x)) x else as.double(x), na.rm = na.rm) : 
  is.atomic(x) is not TRUE
In addition: Warning message:
In mean.default(x) :

Any help to get me going would be appreciated. I'm also aware my formulas are probably wrong, but I want to focus on at least getting some numbers back before worrying about if they're right or wrong. I will give any more information you need if you ask.

CodePudding user response：

First, I would just have it return a vector, so something like:

  return(c(ts = ts, 
           pval = pval, 
           lwr = ci[1], 
           upr = ci[2]))

Then, you could do something like the following:

res <- t(sapply(1:5000, function(x)ev(rnorm(20,0,1), mu0=0, alt="two-sided")))
head(res)
             ts      pval        lwr       upr
[1,]  1.8754644 1.9245952 -20.160528 20.887336
[2,]  0.8501353 1.5946874 -37.338262 37.693908
[3,]  1.0377915 1.6882478 -32.304612 32.766179
[4,] -1.3111520 0.2046614   2.856899 -3.386813
[5,]  1.5704836 1.8680108 -23.406800 24.171175
[6,]  0.7584952 1.5429994 -40.459755 40.911854

As someone who teaches stats, I don't want to just give the answers to you, but I would encourage you to look at a couple of things. First, make sure that the confidence intervals are calculated in the way you intended. Second, make sure that the p-value calculations are as you intend, particularly for the two-sided test. You'll want to make sure that the results you do get make sense.