[Minitab] Determining sample distributions

#1
Hi all,

This is probably a very simple question but I'm new to statistics and don't really know what I'm doing! I have an assignment that asks me to determine the sample distributions of some data. What does this mean and how do I do this in minitab 17? Any help would be greatly appreciated!
Thanks
 

Miner

TS Contributor
#3
Are you trying to identify the type of population distribution (e.g., normal, lognormal, weibull, etc.) from which a sample came?
 
#4
Researchers were interested in the effect of a toxin on an enzyme level produced in cells. 3 different concentrations of toxin were used and the levels of enzyme were measured.

0 20 40
4.1 2.7 3.2
4.1 3.9 3.2
3.4 3.2 3.4
4..2 3.0 3.4
3.0 3.3 2.8
3.4 3.3 2.8
2.6 3.6 3.8
3.5 3.6 2.7
3.7 4.0 3.6
3.3 4.2 3.5
3.7 3.3 3.7

Determine the sample distributions and use an appropriate statistical test to determine the effects of the toxin on enzyme levels.
 
#5
That's what I understood by that question but I'm not really sure how to determine that, I'm only familiar with normal distributions
 

hlsmith

Omega Contributor
#6
It would also help us if you told us what level of course this was and what you have covered (e.g, ttests, correlations, any non-parametric tests?).
 

Miner

TS Contributor
#7
They may be asking you to verify normality assumptions then selecting an appropriate test (i.e., parametric vs. non-parametric) based upon whether the assumptions have been met.
 

noetsi

Fortran must die
#8
The common way you determine what a distribution is would be to try qq plots for specific distributions and stop when you find one that matches :p But I am sure that is not what a class would teach.
 

hlsmith

Omega Contributor
#9
Good point about qq-plots noetsi. I haven't played with data, but they only have 10 units. Also, I love histograms to visually compare measures centrality and dispersion between groups. But once again, your approach will be dictated by what you have covered in your course.
 

noetsi

Fortran must die
#10
What hlsmith said about 'what you have covered in your class" is key. If you start talking about qq plots and you have not covered it in class (which is likely in an intro course) then your professor is going to want to know where you heard about it :p
 
#11
I'm doing a Masters in biomedical science so this is part of preparing us to analyse our own results when we carry out our research project. We have not covered any statistics at all we have just been set questions and asked to research statistics to find out how to answer them.
 

hlsmith

Omega Contributor
#12
Sounds like a reverse classroom anti-teaching method, just kidding. If I was given that I would check normality assumption for ttest/ANOVA and then investigate non-parametrics. Though, if you need to be pairwise comparisons between groups you should look up how to control for false discovery from repeat comparisons.


Good luck.
 
#13
Yeah would be really useful to have even some basic teaching on this but oh well! I'll give that a go and hope for the best, I figure as long as I can justify why I've done what I've done I will still get some marks. Thanks for the suggestions
 

rogojel

TS Contributor
#14
Hi all,

This is probably a very simple question but I'm new to statistics and don't really know what I'm doing! I have an assignment that asks me to determine the sample distributions of some data. What does this mean and how do I do this in minitab 17? Any help would be greatly appreciated!
Thanks
Hi,
under the Statiscs/Quality Tools in Minitab you have a menu item called Individual distribution identification This will give you the best guess for the distribution.
If you only want a normality test, under Statistics/first menu point (standard statistics or simple statistics..something you have a normality test, complete with qq plots and all

If you only need this to decide which test to use, then the normality test is sufficient imo.
regards
 

Miner

TS Contributor
#16
If you are trying to identify an unknown distribution, rogojel's suggestion is the correct approach. If you want to evaluate the data against a specific distribution, use Graph > Probability Plot then select the distribution that you want to plot your data against.
This is Minitab's version of a Q-Q plot.