Referring to your first question: I would recommend to use the syntax to do your caclulations. Then your model should look similar to this one (ignore the dummy variable names)

GLM pos_1 pos_2 pos_3 pos_4 pos_5 pos_6 pos_7 pos_8 pos_9 pos_10 pos_11 pos_13 BY treatment

/WSFACTOR=concentration 4 Polynomial variety 3 Polynomial

/METHOD=SSTYPE(3)

/EMMEANS=TABLES(treatment) COMPARE ADJ(LSD)

/EMMEANS=TABLES(concentration) COMPARE ADJ(LSD)

/EMMEANS=TABLES(variety) COMPARE ADJ(LSD)

/EMMEANS=TABLES(treatment*concentration)

/EMMEANS=TABLES(treatment*variety)

/EMMEANS=TABLES(concentration*variety)

/EMMEANS=TABLES(treatment*concentration*variety)

/PRINT=DESCRIPTIVE ETASQ

/CRITERIA=ALPHA(.05)

/WSDESIGN=concentration variety concentration*variety

/DESIGN=treatment.

Now change the /EMMEANS=TABLES(treatment*concentration*variety) line to these three lines:

/EMMEANS=TABLES(treatment*concentration*variety) COMPARE (treatment)

/EMMEANS=TABLES(treatment*concentration*variety) COMPARE (concentration)

/EMMEANS=TABLES(treatment*concentration*variety) COMPARE (variety)

SPSS will calculate simple effects analyses for your interaction. If your interaction is not significant, you should discuss if your effect size is of interest. If yes, increase your sample size, if not, then it is how it is. Nothing there.

Regarding to your second question: what do you want to show with the correlations?

First Thank you for answering. For the second question i want to show if there are some relation between parameters .but i dont know if i take every 2 parameters from the same treatment alone  (for exemplecorrelation between protein accumulation and ridcle lengh  for humic acid ) or should i have a anouther correlation in general . Cordially

Dear Maher, still I cant answer your second question, since I do not know if it is important to show the differences between the correlations from the treatments or if you generally want to show that there is a overall correlation. This seems to be a research quesstion and not a methodological one. Ask yourself what do you watn to show/proof with it.

Besides that, why not reporting the overall correlation and the correlations for each treatment? You could also calculate if differences between them are significant.

For the first question I suggest you to use an ANOVA analysis with 3 ways;

For the second question, I think that to buid one correlation for each treatment is correct, but for a simultaneous comparison among treatments you would use a MANOVA analysis.

Best regards,

Dear Maher

Your factorial experiment includes three factors : salt concentrations ( A ) with four levels and varieties ( B ) with three levels , and treatments ( C ) with three levels.

In this case, the ANOVA table will include : The main effect of A , B , and C

the interaction A × B 

the interaction A × C

the interaction B × C

the interaction A × B × C

the experimental error

The first question, the post hoc to compared means of interaction A × B ×C need to add a column in data editor and put 1 , 2 , .... , 36 for the treatment combinations , for example, the attached file

Good Luck

I think that previous answers help a lot.

Hereby, I would like answer to your question:   

"And if the interaction is not significant? How to do ?"

If the interactions is not significant, there is no logic to keep the interaction in the model and you should re-run the model without the interaction. In this way, you can estimate the error with more degree of freedom.

If the interaction is not significant, the main effect can be analysed, otherwise the analysis is a bit more complex.

Best regards,

Roberto

Firstly; your research is factorial experimental design. So, i would like to advise you to use other statistical software; example: Genstat, Minitab, JMP or SAS; which are simple and easy for interpreting. you can adjust whatever interaction u want to see among the factors; and the result will be revealed accordingly.

Dear  Maher,

First, your treatment selection and arrangement need to be clear. You have (1) 4 concentration of salt (0 g/l, 2 g/l, 4g/l, and 6g/l); this is considered as the first factor with four levels; (2) you  have 3 varieties a crop (Tunisian, Maroc, Iran) with three levels as a second factor. The factorial combinations of these two factors alone can make 12 treatments. But, you still have a third factor  with two treatments (salt stressed plant humic acid and the second treatment is to give salt stressed plant calcium), which is not clear. If you consider this as a third factor the treatments will be 24, i.e. factorial combinations of three factors with four, three and two levels each. As to my understanding you have three factors with different levels, not variables.

Then, the experimental design you used needs to be very clear. For instance, randomized complete block design or randomized complete block with split plot. Then, as mentioned above you need to select a proper statistical software to analyze your data. I agree, SAS and other statistical packages are more appropriate than SPSS for experiments like yours, though SPSS can also be used. The choice of a particular design depends on the number of factors and treatments you have selected. For a randomized complete block design, your treatments seem a bit higher. If you want to use a split plot you need to know which factors to be assigned in the main plot and which others in sub-plots. This is very important in terms of variances and significant levels.

To compare the tolerance of varieties for salt stress, you may use single degrees of freedom orthogonal contrast, or principal component analysis. But, first your data arrangement needs to be clear. Regarding correlation analysis, you may need to know the type of correlation analysis you want to perform, as there are different types of correlations. SAS is the best software to get all these outputs in one submission.