matrix

I have a matrix (100000x16) where the columns refer to depths and the rows refer to days and the data is of temperature. Some of the data is faulty i.e. gives readings which are obviously wrong. If every row is a measurement taken every 4 minutes how is it possible to state that if any temperature varies by one degree compared to the next measurement (i.e. the next row) to change it to nan?

cheers

Answer 1

[Robert Cumming]

something like:

data(data(2:end,:)-data(1:end-1,:)>1)=NaN

[ricco]

I see what you trying to do but it doesnt seem to work. This what ive tried to do so far:

Firstly I made every value outside of a specific range as NaN, which works:

Temp(Temp(:,5:end)<0 | Temp(:,5:end)>30) = NaN;

The I try to do as you said:

Temp(Temp(2:end,:)-Temp(1:end-1,:)>1)=NaN;

I dont know if this isnt working because I have NaN in the matrix? But without doing the first stage the process wouldnt work, as I have values up to 9999 which is when the instrument taking the readings malfunctioned.

cheers

[Robert Cumming]

it should still work with NaN in your matrix - when you say it doesn;t seem to work - do you get an error message?

[ricco]

No I dont get an error message, it doesnt change the values i.e. I still get values which are greater than 1 degree difference from the next.

Answer 2

[Walter Roberson]

Robert's suggestion of

data(data(2:end,:)-data(1:end-1,:)>1)=NaN

should be close, but his logical matrix will be one element short, which will lead to difficulties.

Ricco, you need to define specifically: when you have two consecutive values that differ by more than 1 degree, is it the first value that should be set to NaN, or the second value? The choice will determine whether Robert's logical array needs to be padded at the beginning or the end.

Also, there should be an abs() wrapped around the difference calculation.