# How UQ Lab uses experimental data and input marginals to construct pce

Hello,

Thank you for your previous responses, they are much appreciated.

I provided UQLab with 17 data points. The ‘X’ (input) is a matrix with 17 rows and 3 columns. Example:
X = [0.0012 28 1243; 0.0012 28 1253; …]
0.0012 is A
28 is E
1243 is T

The ‘Y’ (output) consists of 17 rows. Example:
Y = [176; 168; …]

There are three input marginals. The first input marginal corresponds to ‘A’, it has a gaussian distribution. The second input marginal corresponds to ‘E’, it has a gaussian distribution. The third input marginal corresponds to ‘T’, it has a uniform distribution.

I change the bounds for T for uniform distribution and run the MATLAB code. I do this for each data point to approximate the correspond Y value at the respective T distribution.

Is this correct? I changed third input marginal to a constant and changed the value for each data point but the result was always the same.

How is the experimental data used in conjunction with the input marginals?

Thank you.