CODEWITHSUNDEEP
matlablinear-regression
Blitzkayir
Blitzkayir

Reputation: 67

How to find a linear trend from a time series?

How would I go about finding a linear trend from a time series?

I know the time series is y = {y1,...yn} or for the first case {1.000000000000000 0.917457418407746 0.683456229182811}. It's assumed that each element yi was sampled at time ti = i.

The polynomial it needs to be fitted to is of degree 1.

What I've attempted is below.

function [ linCoeffs ] = getLinearTrend(y)

   y = [1.000000000000000,0.917457418407746,0.683456229182811];
   x = ;
   linCoeffs = polyfit(y,x,1);

end

It's not much as I'm stumped on what to do for the x value. Normally y would be an input value but for testing I just left it in.

Upvotes: 1

Views: 1764

Answers (1)

mattsilver
mattsilver

Reputation: 4396

Looks like you've provided the arguments to polyfit in the wrong order. The first argument takes the values of the x axis which, in this case, correspond to the sampling times given by i. The second argument takes the y values. So you'd want something like:

function linCoeffs = getLinearTrend(y)

   linCoeffs = polyfit(1:size(y,2), y, 1);

end

which, given your example y, results in:

y = [1.000000000000000,0.917457418407746,0.683456229182811];

getLinearTrend(y)
ans =

  -0.15827   1.18351

Upvotes: 2

Related Questions