Statistical methods for normal distribution
Firoz Ahmed
Thursday, 11 June 2009 17:31 UTC
Hello!
I want to know which statistical method can be used to check the mean of a single variance. Suppose I have carried out an experiment (biological) 4 times. Got different values eg 8.5, 10, 7, 6. How could we know that the average 7.87 is statistically significant (with p-value). That means the conclusion drawn based on these experimental data is reliable.
Updated 11 June 2009 17:33 UTC
-
Replies
-
Hello Firoz, I think this would be covered in a basic statistics textbook. Nature’s checklist for statistical methods is here, if that’s any help.
-
Hi,
Please see this link; may be helpful to youhttp://faculty.vassar.edu/lowry/VassarStats.html
-
Hi Feroz, you can find the significance by using SPSS software which is freely available.
-
thanks to all of you
-
Hi Firoz,
Just in case you are testing the effect of some intervention or the difference between two populations, I would like to suggest you to think of it as a question about the statistical power of your experimental design rather than about the statistical significance of the mean you calculated.
In fact, despite some approaches to the sample size/statistical power calculation can be somewhat similar on the “algebric” appearance to those applied to the statistical significance calculation, the questions and the concepts behind them are quite different, and this may drive very specific choices: I would usually go with the interval confidence methods, but many collegues are by far more familiar with the hypothesis testing-like approach.You may find some easy to use (but quite limited) resources here:
http://www.raosoft.com/samplesize.html
http://hedwig.mgh.harvard.edu/sample_size/size.htmlMore specifically, the question you are asking is probably missing some specifications but I will try to give you an answer: the average is by definition representative of the central tendency of the sample from which you have calculated it (then you will not find any test to address such an issue), but you could choose one approach over another (let’s say a weighted mean over a median or a mode) according to some properties of your distributions (which here looks very simple) and based on your final purposes. Concerning the issue of your sample being representative of not of the population you extracted it from, here we would open a completely new chapter but many authors would suggest that raising the size of sample above n=30 would normally work reasonably fine.
By the way, SPSS is a proprietary software (on sell) and the only thing you legally find online is the trial version (which has a usage time limit and some other limitations). I would rather suggest you to check other solutions available from the opensource (I would stick with R, but it has a bit of a learning curve) or the scientific communinities (there are excelent macros for statistical calculations on MS Excel)… or to prepare to esburse something for a software like SPSS itself or Stata.
I hope this will be of help. Take care and good luck,
Marco
-
