# Basic questions on sampling sizes and basic question on sample means

Discussion in 'Probability' started by 2.7182818284590..., Oct 15, 2010.

1. ### 2.7182818284590...Guest

I work in a lab where we collect ~200 data points in a period. The
data that we collect is blood sugar. OK, here is my VERY BASIC
QUESTION:
1. I am not sure how the prob. curve is shaped, but should we simply
take a sample size of 30 for quality control processes?

2. What are the governing equations describing all this?

Finally, here is a very different question that I'm interested in:
Suppose that the average height of men is 70" with a STDDEV of 2.5".

Now, suppose that I collect 2 men per sample, and I measure their
sizes. What would be the STDDEV of this statistical test? I realize
that the STDDEV would be quite smaller, but what are the governing
equations here?

2.7182818284590..., Oct 15, 2010

2. ### Jasen BettsGuest

depends what you are going to do with them.
I think you are looking for the chi-square test, or may atleast find
http://en.wikipedia.org/wiki/Chi-square_test#Chi-square_test_for_variance_in_a_normal_population

Jasen Betts, Oct 17, 2010

3. ### Peter WebbGuest

Ë™ÊžÉ”Ä±É¹Ê‡ ÇÉ”Ä±u

Peter Webb, Oct 17, 2010
4. ### 2.7182818284590...Guest

Hello,

In our blood sugar lab, we are going to take the mean and the STDDEV
for quality control purposes. We are sure that if the average sugar
level of patients is 100 with a small STDDEV, and one random week, the
average is 125, we know that there is something wrong in the way we
are taking data or there is a public health issue at hand. Most
likely, however, there is something wrong with the way we are
gathering data.

My big question are these:
1. Basically, how do we statistically sample all this? IE, do we use
sample sizes of 30 every week? How many samples do we need to be
within +/- 5% of the mean.

For example, suppose that the average is 100, STDDEV is 10, and the
total population is ~200. How many samples do we need to analyze for
a margin of error to be 5%?

2. If we can put all this data into Excel, what would be the
DISADVANTAGES of simply analyzing ALL 200 weekly data points?

2.7182818284590..., Oct 17, 2010
5. ### Jasen BettsGuest

you need all of them to be certain to be withing 5% of the mean,

it appears that you are not intending to re-test the samples but
instead to just re-process the results, so why not put all of them
into the computation?
margin of error is an egineering term, in statistics the best you can
do is a confidence interval, a confidence interval gets you a range
and a probability.

if the test says for examply you're 98% sure that the mean is between 97
and 103 that's probably a good result but if it comes out 98% rpobability
that it's between 95 and 99 you may want to investigate.

the chi square test will tell you if the sample mean (mean of eg. 30
measurements) is likely to be consistent with a population mean of 100.
If you don't need to copy the data by hand I can see no disadvantages.

If the data is in a database I would ask the database to give me
the mean rather than copying the data and then processing it in excel.

I was of the impression that laboratory tests were calibrated by using
a standard solution of known concentration, why not confirm your
processes that way?

Jasen Betts, Oct 18, 2010
6. ### 2.7182818284590...Guest

Excellent answer my friend. I very much appreciate this.

2.7182818284590..., Oct 20, 2010