Convert negative values to positive values in the graph

I have this graph:

The question is, can I convert those negative values to positive? I mean, so that all those spikes in the graph will go upwards instead of go downwards. I mean it will go something like this:

Thank you.

Answer 1

[the cyclist]

Yes, but the answer depends on what exactly you have. Do you have

• just the image
• the MATLAB figure file, but not the underlying data that made the plot
• the underlying data

If you have the underlying data, just use abs to take the absolute value of the variable on the y axis, and re-plot.

[Edoardo]

I have used abs before, but it doesn't help much. The graph changed, but still leave negative values there. I prefer not to leave the data here, since this is for my research. FYI, the data is the edf file used in EEG scanning

[the cyclist]

Can you share some made-up data and some code that generates a plot like the one you are showing? We can't really help with a problem that you can't show us.

I hope you understand that the output of the abs() is never going to be negative.

[Edoardo]

Maybe I will use this as an example: Read EDF File

tt = edfread('example.edf');
info = edfinfo('example.edf');
fs = info.NumSamples/seconds(info.DataRecordDuration);
recnum = 5;
signum = 2;
t = (0:info.NumSamples(signum)-1)/fs(signum);
y = tt.(signum){recnum};
plot(t,y)

[the cyclist]

If all you want to do is for your negative values to be positive, then abs() obviously does the job.

tt = edfread('example.edf');
info = edfinfo('example.edf');
fs = info.NumSamples/seconds(info.DataRecordDuration);
recnum = 5;
signum = 2;
t = (0:info.NumSamples(signum)-1)/fs(signum);
y = tt.(signum){recnum};
plot(t,abs(y)) % All I did was add abs() here

But as @埃博拉酱 has said, the correct solution really depends on the meaning of the negative values.

Answer 2

[埃博拉酱]

It depends on the meaning of negative values and how they are generated.

From your graph, your data appears to be oscillating with a mean of 0, which may be due to normalization. In this case, negative values make sense, and it is recommended to use a method similar to Data-min(Data(:)) so that the minimum value becomes 0.

If you want to calculate a statistic that is not possible to be negative by definition, such as power, you should usually square the data.

If you are very sure that negative values are meaningless error values, it is recommended to use interpolation to change negative values to the average of positive values on both sides.

I can't quite agree with @the cyclist. When dealing with natural measurements like EEG, it rarely makes sense to use abs because it is continuous but non-differentiable at zero, a feature that is too weird in the macroscopic physical world.