Your question is not quite clear but let me try to answer it. If the second develop from the first, you have to conduct causality tests because correlation does not imply causation. In this case, the granger causality test and cointegration tests will be useful. If they measure the same thing, you still need other tests because correlation has its own limitation. You can try relationship tests like rmse, mse to know how close the measurements are or cross correlation tests.

Dear Ayu, Like Samuel T O, I'm not certain I interpret your question right, but I think you have two methods both used to measure the same "thing" of some (experimental) units, resulting in values x and y, say. You want to know how x and y are related.

First consider possible systematic differences. If they essentially measure the same thing, and the "thing" is substantially varying between units, you will observe that x and y are strongly correlated. That is trivial and not very interesting, hence not at all enough. The following is more informative: Consider the difference x-y (or the ratio x/y or the log of the ratio) and plot it against the average (x+y)/2 (or the geometric average), which is a reasonable measure of the object's underlying value (its "thing"), and which is approximately uncorrelated (unrelated) with x-y.  Then you will see if x and y are systematically different and if the difference depends on the "thing" measured. Go on by plotting x-y against other characteristics of potential interest, to investigate if x-y shows any dependence on them, for example (depending on application): time, concentration of another substance, different instruments or persons making the measurements, instrument calibration, different handling of the units, etc. It is up to you to judge what is of relevance in your application. 

When you are clear about systematic differences (or lack of difference), it is of interest to consider if the methods have different precision. Then you need replicated measurements on the same unit, to see and compare their variation, for x and for y.

Best wishes,

Rolf S

If you feel theoretically, the two are correlated, you can test paired t-test to test the significance of the two measurements. otherwise you can use independent t- test for two samples.

Rajiv

Ayu, unfortunately, the comment by Rajiv is misleading in his formulation. Don't ever use two-sample t-test. If the two measurements are empirically positively correlated, the two-sample t-test is wrong, and if they are not at all correlated, this test and the whole situation is not meaningful, because the two measurements are not in any sense measuring the same thing. 

Rolf S

dear all,

thank you for the suggestion, and I am sorry if my question did not clear enough to understand. My two measurement have the same term, but different in elaboration of the construct. The first construct have tree dimensions and the second one have five dimensions. And I want to know if this construct could use interchangeable each other.

Ayu

Are you talking about alternative form of the same construct? If yes then you should go for alternative form reliability of the construct. 

Ayu, I'm a little disturbed by not understanding what you mean by "have the same term", but leaving that aside, if you want to find out if the two measurements give more or less identical results, and you have tried them on the same items, I again suggest you use the procedure I outlined in my first reply.

Rolf S