How to change selected data which would give me the line of best fit (R squared close to 1)?

I have a range of data

x = -Inf + 0.00000000000000i
-5.94839731081433 + 0.00000000000000i
-4.40441244315510 + 0.00000000000000i
-5.01054824672844 + 0.00000000000000i
-5.79424663098060 + 0.00000000000000i
-4.91523806692085 + 0.00000000000000i
-2.61442132098323 + 3.14159265358979i
-3.99894733505125 + 3.14159265358979i
-2.61265297392781 + 3.14159265358979i
-4.00012588556654 + 0.00000000000000i
-5.20409868989224 + 0.00000000000000i

y = -3.45375423084893
-3.45376083103347
-3.45405094838298
-3.45428146199255
-3.45431857619349
-3.45456107498460
-3.45493413618944
-3.45513888357560
-3.45551262399935
-3.45574790889255
-3.45589974019472

There are 2 things I want to find out.

1. How to set the values of complex numbers in the array x to NaN? 2. How to select the range of data with linear trend and with best fit (R squared close to zero)?

[Walter Roberson]

Unfortunately, the two things that you want to find out went missing in the Question ?

[Joe Abraham]

sorry about that ! question modified !!

Answer 1

[Richard Brown]

For the first part

x(logical(imag(x))) = nan;

For the second, your data points are all over the show. Any two will fit a straight line pretty well though :)

More seriously, try either a 1-norm best fit, or a method like RANSAC if you know that a subset of the data should fit a line and the rest are outliers.