filling in blank cells with nearest number

Hi there,

Would anybody be able to help me with some code to 'fill in some data'.

I have a column of data with '1's and '0's with blank cell in between. I'm trying to fill in the blank cells with the nearest number.

Greatly appreciate any help. I have 105,841 rows.

Here is an example:

[Wayne King]

how are these "blank cells" rendered in MATLAB as NaN?

[John]

Hello Wayne, thanks for your reply. Yes they are rendered as NaN.

Regards

Answer 1

[Wayne King]

The question I'm not sure of from your post is how you want to deal with

1 NaN 0

0 NaN 1

Do you want the above to be assigned 1?

I'll assume that you have at most 1 NaN between numbers.

There are many ways to do this

x = [0 0 0 NaN 0 1 1 NaN 0];
indexvec = 1:length(x);
idx = isnan(x);
indexvec = indexvec(idx>0);
x(indexvec) = max(x(indexvec-1),x(indexvec+1));

The above code assigns a 1 to both

 1 NaN 0

and

 0 NaN 1

It also assigns 0 to

 0 NaN 0

and 1 to

 1 NaN 1

[John]

Hello Wayne,

Thanks for your help.I have 105,841 rows and there could be up a few hundred NaNs between the numbers. Would there be anyway to perhaps to find each set of NaNs and then check either side for the 1s and 0s and then overwrite the NaNs. I appreciate this is an unusual problem and am grateful for any help.

This is what it looks like:

[Wayne King]

It's not that unusual, but please define what you want when a long list of NaNs is buffeted by different values, for instance, in your screen shot, should values 6-10 be assigned a 1 or 0, what about 5-8?

[John]

Hello, my apologies for not defining what I want clearly. What I am looking for is, when a long list of NaNs is buffeted by different values, I want to make half of them 1s and the other half 0s.

For example 5-8 are buffeted by a 1 and 0. I need to assign 5,6 as 0 and 7,8 as 1. This example is easier as there are 4 (an even) number of NaNs.

For example with an uneven number of NaNs, 16-20, are buffeted by a 1 and 0. I need to assign 16,17,18 as 1 and 19,20 as 0. The rule is the case of an uneven number of NaNs is to let middle cell be a 1.

Thank you

[John]

Hello Wayne,

Sorry for bothering you, I'm just wondering if you had time to consider my problem any further.

Kind Regards