Dear Nguyen

I think it will be better to run 2 way ANOVA as T-test is done for comparing 2 samples. You can run 1 way ANOVA to test difference among the samples and 2 way for samples and layers also.

Regards

 Dear Dr Tho Nguyen, 

i presume that the question you want to ask is " is there a statistical significance between mean ( Chemical parameter) , using a depth interval of  10 cm?"

So the independant variable in your exepriment is the soil depth ( 10 cm), and the dependant variable  or the response is the chemical parameter, so you can perform the classical one way ANOVA, and then compare for each borehole , the mean of each parameter along the vertical axis. you can use,  depending on the software that you have, i use Stastica , and the  Newman and Keuls test is useful for comparing the mean. 

Best regards Louadj yacine.

The experimental design you have mentioned is a single factor full factorial design with multiple responses. It is possible to run a one way ANOVA for one response variable at a time or two way ANOVA for all the responses together. This will give the answer for a presence of significance difference among the bore holes in each response variable.

Further, it is possible to run Tukey's honest pairwise multiple comparison, LSD or Duncans multiple comparisons or the significant differences among one response variable of one bore hole against the same response variable of another bore hole and so on.

It is also possible to visualize the differences and simillarities among the response variables and the bore holes through Principal component and principal coordinate analysis in a two or three-dimensional space. 

If you have any standard values for each response variable it is posssible to do t-test. In SAS there is a graph which plots the bar graphs vertically and shows the standard and the confidence levels as parallel horizontal lines 

In addition to the above mentioned test Cochran's test can be performed

Thank you for this question, I had exactly the same, and I was not sure the samples could be considered as independant (and in the same time, they are not as one sample on which two treatments have been tested). According to the responses, I have the impression it is not shocking to use ANOVAs or Kruskal-Wallis.

If the values are not normally distributed then you have to run the Kruskal-Wallis test, but if they are normally distributed then 2 way-ANOVA is OK.

If you are interested to test the difference of one chemical parameter like potassium among six soil horizens, then one-way ANOVA is suitable.