Hello,
Can any one tell me what sample size should I use for mechanical properties testing of metal foam? Some have followed sample size of say width (10 times pore size) while some say the width should be at least 7 times the pore size.
Can I use width of 7 times the pore size for compressive testing ?
Dear
In terms of statistics two type of sample,
Large sample is > or = 30 and small sample is < 30.
For two kinds of sample we use two different type of theory. So it's not a fact about the sample but it fact that the theory depends on the sample size as it large or small.
Thank You.
It depends on the precision necessary for the estimate. There are several related ways to work this. A simple form of this question is to determine the maximum difference where you are happy to decide that there is no difference. So if pore size of 5 takes 2 kg/cm2 to deform, while a pore size of 6 takes 2.005 kg/cm2, is this a meaningful difference in the context of your problem?
Another way to ask the question is to understand that there is an inherent variability in your data. If you measure pore size for 1000 pores in your sample they will not all be exactly 5.0000000000000 microns in diameter. Possibly your manufacturing technique is very consistent, and pore size differs by only a few molecules and that is below the ability of your equipment to measure. But there will be variability. To this variability are a range of other factors. Small changes in temperature, how you handle the equipment, micro impurities in your sample, and the list goes on. To this is added variability due to sampling. If I take 4 samples from a Normal distribution and 4 samples from a uniform distribution, you will not be able to tell me which set of samples came from which distribution. As you take more samples, it becomes more likely that your estimate of mean and variance will accurately reflect the population mean and variance. Play with a few random number generators to work this out.
There are some web resources that will provide sample size estimates. GPower is one such tool. However, they often want information that is not available.
Another option is to run your experiment and take as many measurements as possible. Run the data analysis and figure out how many samples you should have taken. Redo the experiment if necessary. This is not always a good solution because sometimes you get a very different answer when you rerun the experiment. However, if you get a different answer, then maybe the project needs more planning.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size