Missing Observations error when running regression

I am using MATLAB to complete some regression analysis,

Say I have 40 variables, 35 have the full dataset, monthly data from 1950-2010,

However, some have missing values at the start of the data set,

For example, 36 may only have 1969-2010, 27 1952-2010,

I tried to replace all blanks with a NaN and got the error message,

"Error using svd Input to SVD must not contain NaN"

Is there something I can do?

[Sam P]

To be more clear,

I want to just run the shorter regression,

So if the data is missing,

I want to run y =beta(x),

If x36 only starts from 1952 I want the output of the above equation from 1952

Answer 1

[Shrestha Kumar]

Hi,

You can replace NaN with the mean of the data in the column(if it is a numerical value else take the mode). Also, NaN values can be present in the test data, so take care of that too.