Search All of the Math Forum:

Views expressed in these public forums are not endorsed by NCTM or The Math Forum.

Notice: We are no longer accepting new posts, but the forums will continue to be readable.

Topic: resampleing for one sample t-test
Replies: 0

 Pat Ballew Posts: 356 Registered: 12/3/04
resampleing for one sample t-test
Posted: Mar 3, 1997 3:56 PM

I apoligize if this is a repost... it never showed up on my file

----------------------------------------------------------------------

I have a question about resampleing that one of my students came up
with and I hope someone on the list can help me understand and be able
to answer his question. Perhaps Prof. Buce is "listening" out there.
Using data from the death age of males (data shown below,I think I
got this data from the DASL data base but don't remember for sure) The
actual output from the program is below. In brief, his approach was
to use the program RESAMPLEING STATS to draw samples of 18 from the
data values without replacement and find the mean of the sample. This
was done 1000 times and then the 2.5 and 97.5 percentile values were
found for these results.
His observation was that in most situations the confidence interval
found by resampleing was a little narrower than the 95% CI given by
mini-tab. I had him nest the test in a larger loop and repeat it 50
times and find the mean of the lower and upper Limits of the CI and
that is what he did. The CI created by the average of Lows to Avg of
highs from resampleing was (68.44 ... 80.167) as compared to the
minitab results of (67.66...80.78). Taking the half interval width of
the resampleing averages CI and dividing by the SE of the mean given
by mini-tab we estimated that the resampleing averages would be about
a 92% CI for the t-test, but with a slightly different mean... NOW
the question is WHY?? I know this is a small difference, but it was
consistently true (we ran this several times) that the resampleing
means interval fell Inside the 95% CI by mini-tab.
Is this a difference in the accuracy of the tests? .. Something
related to the use of pseudo-random generation of numbers?... or have
we misunderstood the resampleing approach to the t-test question?

The results are below, and include the program file for those who Know
the software.
I am inspired by the understanding their questions show, but
frustrated by my inability to answer. I feel undereducated for the
task I have assumed. Thanks in advance to anyone who can help.

Pat Ballew
Misawa, Japan

__________________________________________________________________
results file
-----------------------------------------------------------------
Start execution.

'simulated t-test for one sample 95% conf interval ' minitab gives
limits of 67.66 to 80.78
'norm plot p-value is .478
'data from minitab sample age dagem
copy (55 60 62 71 74 79 84 90 98 61 70 58 70 93 77 62 82 90) a
repeat 50
clear z
repeat 1000
sample 18 a b
mean b c
score c z
end
percentile z (2.5 97.5) conint95
min conint95 lv
max conint95 hv
score lv zl
score hv zh
end
mean lv meanlv
mean hv meanhv
print meanlv
print meanhv

MEANLV = 68.444

MEANHV = 80.167

Successful execution. (1485.3 seconds)