Your friend is interested in generating a bootstrap sample from some original sample by sampling its observation indices. But you observe that he types sample(n,n,replace=TRUE) in R (where n is the original sample size). You argue that he should type sample(n,n,replace=FALSE) instead. Who is correct? Both operations are acceptable and give the desired result. The replace input argument here is redundant and would only matter if we are carrying out cross-validation. You are both wrong. You need to use more information in order to be able to generate a legitimate sample. You are right. The sampling has to be done without replacement. Your friend is right. Sampling without replacement will give identical estimates with respect to the original sample.
Your friend is interested in generating a bootstrap sample from some original sample by sampling its observation indices. But you observe that he types sample(n,n,replace=TRUE) in R (where n is the original sample size). You argue that he should type sample(n,n,replace=FALSE) instead. Who is correct?
Both operations are acceptable and give the desired result. The replace input argument here is redundant and would only matter if we are carrying out cross-validation. |
||
You are both wrong. You need to use more information in order to be able to generate a legitimate sample. |
||
You are right. The sampling has to be done without replacement. |
||
Your friend is right. Sampling without replacement will give identical estimates with respect to the original sample. |
Step by step
Solved in 2 steps