Removing strings or NaNs from data to be plotted

Hi, I have txt files which represent results of measurements of objects in flight and occasionally an object will not be present, as a feature this is of as much interest as the objects we get data for. So the txt file will show the missing objects with a string of letters, all other fields in the file being numbers (except for the headers). I have tried for days to replace the strings with an empty cell, in a cell array, but this cannot be plotted directly from the cell array. When i convert the Cellarray to doubles the blanks are filled by NaNs which I cannot plot either. I am sure I am missing something obvious but I have been reading all the documentation online and tried to tweak solutions to other peoples questions to no avail, I always end up with something that only does half the job I want. Any ideas?

Best regards

[Stephen Devlin]

Hi, I was not asking for a "This question is: * Do it for me * Not likely to be useful to future visitors * Too open-ended", I was asking for general advice. I have converted my txt files into cells, structs,doubles and always there is some problem in one of the arrays that I can only solve in a different type of array. Specifically I wanted to know if there was a way to have a double array where blank/empty cells were preserved and later a plot of that data would show an empty space, from my reading of the documentation that is not possible as when the cell data is converted into a double array the double array is resized and the data is moved to fill any empty spaces, that is of no use to me in this instance. I have cells where the data is exactly how I want it, but I cannot plot it from the cell columns.

Answer 1

[dpb]

Well, ideas would depend upon what you want -- plot and friends will gladly handle NaN; they even show where they are located if you use a line by making breaks at those points. If you were to want to make some different sort of mark at those locations, one option would be to first plot normally; then replace the NaN values with (say) an interpolated value and use scatter to add those as points to the preceding plot. Any number of alternatives could be imagined; just depends on what you're trying to accomplish.

[Stephen Devlin]

hi dpb, Plot is not something I have used much, I tend to use scatter and stem so I am assuming these two are similar. It wouldn't be necessary to have a line with a gap, if the plot shows that there is no value for that x index that would be fine. I found a very long winded way to do it but it meant pulling each column out of the array and dealing with it as an individual vector, I hope to get a solution where I can just index an array without having to do that (but it does work so I have that to fall back on, albeit a fudge solution).

Best regards Steve

[dpb]

No difference really between scatter and plot excepting for some reason scatter isn't vectorized() to accept arrays as is *plot--never did understand that limitation. But, you can easily-enough do a scatterplot with plot; just set the linestyle as a marker. But if you arrange your data by column, then it's easy-peasy with only a line or two of code at most--

Above was made with:

>> dat=randn(10,3); dat(randi(30,5,1))=nan;  % some data w/ missing
>> plot(dat,'*-','markersize',10);
>> legend(num2str([1:3].','Column %d'),'location','best')

Note the broken lines w/ markers at the points. Simply replace the linestyle string '*-' with '*' to have markers only; the difficulty with plot vis a vis scatter would be to have the different colors by data value or the like; the lines are individual objects with their own properties of one linestyle, etc., per line instead of individual points.

() There's *plotmatrix but it has the annoying characteristic in that it only produces an NxM array of axes rather than being able to put the columns on a single axes.

[Stephen Devlin]

Hi dpb, I will have a look at this today, thank you for your help.

[dpb]

Well, if you have "issues", supply more details of what the effect(s) you're trying to create are and undoubtedly others may well have better solutions. Oh, there's also always the File Exchange -- somebody else may have already built an enhanced mousetrap for the type of plot you're envisioning, particularly if it is a type commonly used in some area of data analysis/presentation...