Jayachandran, 

Samples may be water quality parameters or species names.

Variable may be sites/some other thimgs.

Select New data..then Sample data...then Unknown/other...Click Ok. Dont forget to transform (square root).

And everything is up to ur data how to design.

All the best!

Regards, 

Dr. Tenjing

But how to selected parameters for analysis?

I mean we have have do, ph, salinity , temperature etc  (100 parameters) as environmental parameters.

How can I selected or predict most important variables  influencing our study area or our organisms?

As you know we cant load so many parameters to PCA graph it may be not looking good?

It does not matter. In one case, u can perform cluster analysis and nMDS for clear-cut. 

I mean to say Temp., pH, salinity, DO, and nutrients.

But u should know that the u may perform Pearson's correlation matrix.

you can also work on SPSS........

Hi jay.

Why do you need to reduce parameters for pca?..isnt one point of pca to produce principal components' and define strength (correlation) of variables to axes.....Provided data is 'cleaned', non autocorrelated and transformed appropriately. You could run a few trials. 

Bioenv bv-step is better than the BEST routine for determining biotic/abiotic relationships for large datasets. 

Hope ive understood query correcly.

Barry

Jayachandran,

You may use PC scores to reduce variables (i.e. the correlated ones). firstly, you should do a PCA on your parameters using predetermined number of PCs (lets say N= 10 PCs). Then check to see how many PCs are needed to describe 100% cumulative %variation (for example the first 5 out of 10 PCs). In each PC (1st to 5th) choose the variable with the highest score (irrespective of its positive or negative sign) as the most important variable. Since PCs are orthogonal in the PCA, selected variables will be completely independent (non-correlated).

Hope this helps you

Amir

Thankyou Amir

Amir Ghazilou, what is the relevance of Eigenvalues in that

https://www.researchgate.net/post/PCA_analysis_is_done_with_mean_values_of_environmental_variables_or_raw_values

I have a similar situation with my dataset and I wanted to get some help. So in my case, I ran a PCA with these 7 environmental variables velocity, depth, width, slope, distance from upstream, and temperature. I am trying to relate them with the fish abundance data across many sites. My analysis results suggest

Eigenvalues

PC Eigenvalues %Variation Cum.%Variation

1 2.36 34.6 34.6

2 1.75 25.6 60.2

3 0.884 12.9 73.2

4 0.607 8.9 82.0

5 0.523 7.6 89.7

Eigenvectors

(Coefficients in the linear combinations of variables making up PC's)

Variable PC1 PC2 PC3 PC4 PC5

Velocity 0.206 -0.502 0.288 0.237 -0.626

Depth -0.461 -0.172 -0.036 0.814 0.152

Width -0.444 -0.410 -0.022 -0.451 0.154

Slope 0.389 -0.439 0.114 -0.092 -0.015

Ups distance -0.488 -0.382 0.107 -0.229 -0.019

Elevation 0.393 -0.407 -0.042 0.127 0.711

Temp 0.065 -0.215 -0.943 0.009 -0.236

In this case then should I infer that environmental variables influencing fish abundance in terms of importance are 1) upstream distance 2) velocity 3) temp 4) depth and 5) elevation? I also wonder that I haven't checked them for spatial autocorrelation, is there a way to check for this in PRIMER-E?