8 replies to this topic

[conan_kaya]

Hi all,

Can anybody recommend me a statistical tool for finding correlation.

[breizh]

Good day ,
Let you try Excel , there is a function called "linest " which should support your query .
Regards
Breizh

[conan_kaya]

can i know like how to find this "linest" function in excel?

[Trent F Guidry]

It really depends on what you mean by finding a correlation.

For a simple linear regression, that can be done in Excel.

Some nonlinear equations, such as the Clausius Clapeyron equation can be rearranged so that they can then be linearly regressed. While this sort of rearrangement isn’t really a true regression of the original equation, it is usually fairly close.

If you have to do a nonlinear regression, such as for the Antoine correlation for vapor pressure, then that gets much more involved. One of the more popular algorithms for doing that is the Levenberg Marquardt algorithm. The Gauss Newton algorithm can also sometimes work, but it tends to have pretty severe stability issues. Another approach is to create an equation of the residual errors and then use an optimization routine to minimize that function. I have seen that approach taken using quasi Newton algorithms such as the Davidson Fletcher Powell and the BFGS algorithm.

[breizh]

can i know like how to find this "linest" function in excel?

On excel , let you use the wizard (?) and search for LINEST , you will find the description of the function and examples .
regards
Breizh

[gvdlans]

You can also used specialized software. I've used Tablecurve in the past. You can download a free 30 day trial version here: http://download.cnet...4-10308067.html

[wil ma]

...or you can use datafit www.oakdaleengr.com wich also has a time limited demo version

[conan_kaya]

...or you can use datafit www.oakdaleengr.com wich also has a time limited demo version

For example, lets say i have the feed sulphur weight percent (S), the CCS 50% cut point temperature and CCS 100% cut poinnt temperature, is there any tools that i can use to try to form an equation using all these 3 variables to predist the weight percentage of sulphur in the product CCS?

[wil ma]

to form an equation you need a significant amount of (experimental) data.
the equation will only be valid (reliable) in the range of your data.
if there are no other parameters disturbing/influencing the %S in the product CCS then you can make a equation with one of those tools.