How I do a Linear extrapolation?

Hi, I have a matrix with 160 numbers, but the first values is nan. How I complete this matrix? Anyone help me? I try the -- interp1 --- but I have only a matrix, I do not have a vector to realize the extrapolation.

Thank you for your attention. Best wishes, Paulo Oliveira

[José-Luis]

Extrapolation is tricky business. Depending on the sort of data different extrapolation methods might be warranted. The closest value? A linear interpolation from the first two points. A cuadratic function using the last three points? A spline of n degree?

Extrapolation caused the crash of the Challenger.

[Paulo Oliveira]

I understand your questions, but as I said, I have 160 values, but 40 are NAN, so I have a 140 values to support,for you what the best way to do the data extrapolation?

Answer 1

[Star Strider]

This works:

% Original vector: 
x = 1:10

% Replace the first 4 elements with ‘NaN’: 
x(1:4) = NaN

% Create empty elements for the ‘NaN’ elements: 
x(isnan(x)) = []

% Extrapolate to fill the empty elements: 
xm = interp1(x, -3:0, 'linear', 'extrap')

produces:

xm =

       1.     2     3     4

[Paulo Oliveira]

Hi Star Strider,

I try with my matrix and works like I want. :)

Thank you for your attention.

Best wishes, Paulo Oliveira

[Star Strider]

My pleasure!

[Anwaar Alghamdi]

@Star Strider Does this work if I have missing hourly temperature values? (not NaN but 999)

One or two missing values is fine but if a have a whole day missing? Can this preserve the behavior of hourly temprature profile within the day?

[Star Strider]

@Anwaar Alghamdi — I just now saw this.

It depends on what the data are and where the values are that are NaN. The fillmissing function would likely be most appropriate.