What do you mean "how to calculate"? If you have a pre-variable and a post-variable it's analyse-->Compare Means-->Paired-Samples T Test, define the pairs, Paste and run the syntax.

Or what is your question?

With a paired samples t test, it is also worth being aware of the test's assumptions which are that the difference between the two variables is normally distributed. The test will work robustly in some other circumstances. You can test normality using the Analyze - Descriptive Statistics - Explore function, selecting Options and Normality plots with tests. Concentrate on the Shapiro Wilk test for sample sizes under about 2000. If your data is not normally distributed then consider using the Wilcoxon signed ranks test which is available under Analyze - Nonparametric test - Repeated samples.

Mr. Tiwari,

I support the question asked by Mr. Duesing. If you need to actually calculate your parameters, then the advice given by Mr. Duesing is in order.

If you need a path or a procedure to perform a T-test in SPSS program, use the advice that Mr. Samuels gave you. First, run normality test distribution. If you notice that your distribution in one variable is not normal (for example, if distribution in variable TDS on pre-test is normal, but distribution in variable TDS on post-test is not normal) i would also suggest a Wilcoxon Signed Ranks test, which is equivalent to Paired-Samples T-test.

If distribution in your variables are normally distributed, it is appropriate to use Paired-Samples T-test, in the manner that Mr. Duesing explained to you.

Just keep in mind that the scores in your variables are need to be on interval (at least) or ratio scale to even concider performing Paired-Samples T-test.

Good luck with your research.

Dear, first make two variables of each water quality parameter in SPSS (e.g., pre.Ca and post Ca). then create another variable that is difference of pre and post values of each parameter. 

now check the normality of data just by using new variable (difference of pre and post score). 

You can test normality using the Analyze - Descriptive Statistics - Explore function, selecting Options and Normality plots with tests. Concentrate on the Shapiro Wilk test for sample sizes under about 2000. If your data is not normally distributed then consider using the Wilcoxon signed ranks test which is available under Analyze - Nonparametric test - Repeated samples. 

if data is normally distributed then go to Analyze - compare means then paired sample t test.

Thanks Rainer Sir your valuable suggestions. I have water quality parameters data for both the seaons, But i dont how to prepare excel sheet for paired t test ? or how to fill data in data view or variable view  in SPSS?

 Thanks Peter Samuels for ur valuable suggestion.

Thanks Boban

Thanks Waqas Latif

Dear Ashwani!

In principle, you don't need to prepare your data in excel first. You can insert your data directly into SPSS, if you want to. On the other hand, if you have your data already stored in excel, then it could be worthwhile to prepare them first and save the edited data in a separate file. It is possible to import data from excel or .txt. files into SPSS.

Nonetheless, your data should be structured as follows: rows are cases (persons or maybe in your case the different sites where you collected data) and colums are the variables. So, for example, if you collected pH values pre and post Monsoon at 50 different places/sites, you should have a 50x2 matrix. If you would have collected the ten parameters mentioned by you (Ca Mg, Na, K, pH, EC, TDS , F, Cl, NO3), you would have a 50x20(10x2 for pre post) matrix.

As mentioned, now you could easily conduct the t-tests for eacht pair. If you want to check the normality of all difference scores, which has been proposed by the others and is an assumption for dependent t-test, you have to calculate the difference scores by hand, i.e. with the compute command. Open a syntax and type for example

COMPUTE pH_dif = pH_pre - pH_post. EXECUTE.

After executing this command, a new variable should appear in your viewer.

Now you can check if this variabel is normally distributed. Alternatively, instead of conducting the dependent t-test with the pre and post variabes, you could now run a one sample t-test against 0 with the dif variable. You will see that this will be the same as the dependent t-test, because that's what the dependent t-test actually does and why the differences have to be normally distributed and not the source variables itself.

Feel free to write a pm, if you have any questions.

Thanks Rainer

Dear Ashwani

I may be wrong but you must be careful while using the Paired vs Independent. The two observation should be related in some fashion for paired t test however, I am not sure, how we can say that water parameter (say pH) at two point of time is related. 

Pl see, theoretical basis also.

Ya pl ensure the assumptions before proceeding to test.

Regards

Rajiv

http://www.sciencepub.net/nature/0709/15_1130_gang_aug_ns0709.pdf

may be the answer of your question

regards

Thanks Rajiv Pandey

Thanks Mostafa