Question

I am reading the data set of Arrhythmia, and some attributes are missing (when this happens, they put ‘?‘). For example:

... 64, -2, ?, 63, 0, ...

How should I treat this example? Is it better to get rid of the unknown values when I am doing some statistical analysis?

How do I remove data containing ‘?‘ in MATLAB?

Answer 1

To convert the string, use TEXTSCAN and STR2DOUBLE, which will replace missing values by NaN (for many statistical analysis, you don’t want to just delete the value, but rather you want to know something’s missing). Most functions of the stats toolbox, as well as e.g. plot can handle NaN.

For example

tmp = textscan('-2,**?**,63,0','%s','delimiter',',')
str2double(tmp{1})
ans =
    -2
   NaN
    63
     0