The analytical procedures firstly, depend on the objective of the study that is whether the comparison is done based on qualitative or quantitative measurements. Secondly it has to be made clear that the researcher is interested in correlation or just association of input and output variables. And lastly how the data was collected that is sampling frame of the study

Its an experimental study..i want to see whether the drugs/substances have hypotensive effect or not. So i am measuring the blood pressure at certain intervals for a duration ..suppose 4 hours.

In order to compare mean values of blood pressure for a single sample, among different time points you can do ANOVA (analysis of variance) for repeated measurements and post-hoc multiple comparisons (e.g bonferroni corrected). In SPSS software this analysis is done by using GLM (general linear models for repeated measurements). This is the "parametric" approach (large samples and only if blood pressure follow the normal distribution). In our case (experimental study), I expect that the sample size is small (n

As Farah Mansuri mentioned the different points determine which analysis method to use...I am just seeing the association of input and output variable and the sample size is less than 20...I went through the statistics book and found that parametric test like student t test can be used..i ll again go through the answers as you have suggested. Thanks.

If the groups are matched hemodynamically at baseline and the distributions are normal, then you can go ahead with student "t"-test. However, for within group variability you need to go either for repeat measures ANOVA / paired "t"-test. In case of asymmteric (non-normal) distributions you need to follow the non-parametric evaluation plan which includes Mann-Whitney U test (like Student "t"-test), Wilcoxon signed rank test (like paired "t"-test) and Friedman test (like Repeat measures anova).

Thanks Varun. But It would be better tell me with example like If I have to test an antiglaucoma drug is better than placebo and I chose two eyes as control and test. so how should I go on .

Are you measuring the IOP?

Depends on the number of eyes you have. There is evidence that two eyes of an individual might have different IOPs, hence baseline (before administrtion of drug) matching of two groups (placebo and case groups) is essential. For this purpose you can use paired "t"-test.

If there is no statistically significant difference between case group and placebo group at baseline then you can compare the two groups at different time intervals using paired "t"-test itself. However, if there is a significant difference at baseline, then you are left with comparing the % change in IOP.

All these assessments are subject to normality of distributions.