The chief difference is replacement. Bootstrapping is a replacement method, and sampling distribution is a randomization method.
The bootstap distribution lies more because of the fact it replaces values.
When the percentile method is true. And have a 95% confidence.
The mean will not be as accurate. One will need to do many tests to find a valid answer.
# A tibble: 1 x 2
`2.5%` `97.5%`
<dbl> <dbl>
1 NA NA
# A tibble: 1 x 1
stat
<dbl>
1 1996.
# A tibble: 1 x 1
mean
<dbl>
1 51049.
# A tibble: 1,000 x 19
# Groups: replicate [1]
replicate state emp_length term homeownership annual_income
<int> <chr> <dbl> <dbl> <chr> <dbl>
1 1 FL NA 36 mortgage 94000
2 1 CT 0 36 mortgage 52000
3 1 OH NA 60 mortgage 70000
4 1 NY 10 36 mortgage 90000
5 1 MI 10 60 mortgage 60000
6 1 NY 1 36 mortgage 52000
7 1 CT 10 36 mortgage 114000
8 1 TX 10 36 mortgage 48000
9 1 OK 8 60 rent 30000
10 1 MO 2 60 rent 47000
# ... with 990 more rows, and 13 more variables: verified_income <chr>,
# debt_to_income <dbl>, total_credit_limit <dbl>,
# total_credit_utilized <dbl>, num_cc_carrying_balance <dbl>,
# loan_purpose <chr>, loan_amount <dbl>, grade <chr>,
# interest_rate <dbl>, public_record_bankrupt <dbl>, loan_status <chr>,
# has_second_income <lgl>, total_income <dbl>
Setting `type = "bootstrap"` in `generate()`.
# A tibble: 1,000 x 2
replicate stat
<int> <dbl>
1 1 52092.
2 2 51876.
3 3 53927.
4 4 52285.
5 5 50170.
6 6 50506.
7 7 51699.
8 8 50902.
9 9 52123.
10 10 51284.
# ... with 990 more rows
# A tibble: 1 x 1
mean
<dbl>
1 51414.
# A tibble: 1 x 2
`2.5%` `97.5%`
<dbl> <dbl>
1 48694. 54275.
# A tibble: 1 x 2
lower upper
<dbl> <dbl>
1 46791. 52401.