Stats Question
Six Sigma – iSixSigma › Forums › Old Forums › General › Stats Question
 This topic has 5 replies, 4 voices, and was last updated 16 years, 4 months ago by Openball.

AuthorPosts

April 7, 2006 at 12:04 pm #43014
I have 107 willow occurences (individual trees). Let’s call these theSampling Units (SUs). I have 8 quantitative environmental/physicalvariables (e.g. elevation, valley width, stream slope, etc.) measuredaround each SU. Therefore, I have 107 rows and 8 columns in my dataset.How can I find out if there exist any differences between SUs based uponthe variables measured?
0April 7, 2006 at 1:47 pm #136093
OpenballParticipant@Openball Include @Openball in your post and this person will
be notified via email.Hi REG,I am a little confused with your question. Do you want to test whether the trees planted at different places are correlated? Basically, what do you have are variables related to the enviornment around the trees. However, these varialbes are like the INPUT, not your response. The relationship among these varialbes depends on how (where) do you choose your samples, right? So, I think what you need is to measure the trees from different aspects. Then, analyze where these trees are similar in someway.
0April 7, 2006 at 2:10 pm #136095
Rajesh MohandasParticipant@RajeshMohandas Include @RajeshMohandas in your post and this person will
be notified via email.Hi Reg,
Run a one way unstacked anova to know the variances between and within them.
Go to minitab – stat – Anova – One way unstacked
Under the responses enter 8 quantitative environmental/physicalvariables (e.g. elevation, valley width, stream slope, etc.) measuredaround each SU.Regards,
Rajesh.0April 7, 2006 at 2:54 pm #136099
ImpecuniousParticipant@Impecunious Include @Impecunious in your post and this person will
be notified via email.Before we jump into analysis like ANOVA (which doesn’t seem appropriate given that quantitative inputs are described), we need to understand what the output is.
In the original inquiry, 8 total columns of data were indicated, and 8 quantative environmental variables were mentioned. Consequently, you can’t compare whether one willow tree is different from another because we apparently don’t have an output measurement for them. If we can get the trees’ height or circumference, for example, then we can understand whether the 8 inputs have some impact on the trees.
I’m curious whether one of these 8 variables is really an output.0April 7, 2006 at 4:00 pm #136106
OpenballParticipant@Openball Include @Openball in your post and this person will
be notified via email.I really don’t think onway ANOVA (unstacked) is suitable for this problem. The data in different columns represents different variable, not different level of one variable. This is an interesting problem..I am still thinking possible solutions.
0April 7, 2006 at 4:24 pm #136107
OpenballParticipant@Openball Include @Openball in your post and this person will
be notified via email.I think I understand you problem now. You want to see whether the environment for willow occurrences is different. Am I right? Of course, you first step shoulb be looking at each individual variable, and see whether there are special patterns exist among them.Then, I propose two possible ways to do the multivariate analysis:1. Use a multivariate T2 chart to plot your data. Although you are not really using the control chart to monitor anything, from the multivariate t2 chart, you can find the dispersion of your data. If all data points are very disperse, that means they are different. 2. Use clustering. Try to cluster you data into multiple groups, and check the distance between the groups. If the distance is large, then it suggests that there is significant differences exist in your data. One more thing, you can even work on your original variable first, use factor analysis or principle component to extract significant factors, then do the above two analysis based on the factors your extracted.
0 
AuthorPosts
The forum ‘General’ is closed to new topics and replies.