Find previous time conditions were met, save it as new marix

I am working with ecology data. Presence/absence data (Y) has been collected for 100 sites (X) over 10 years (Z), on a rotational basis (30 per year). To illusterate: site X(1) could, for example, be surveyed year [1,4,7,10], site X(2) surveyed year [2,3,4,8], site X(3) surveyed [3,5].

To better explain observed presence/absence (Y) in a given year z I would like to use presence/absence data (Y) from the previous time a given site x was surveyed.

I am at a loss where to begin. Any advice/pointers to get me started would be much appreciated.

Many thanks

J

[Bob Thompson]

I don't quite understand how your data is organized. Do you have an array with all the information where rows are different sites, columns are different observations, and sheets are different years?

However you have your data organized, if it is all within Matlab at once then I would use some form of indexing to reference different years of data.

[SBinary]

I don't quite understand how your data is organized. Do you have an array with all the information where rows are different sites, columns are different observations, and sheets are different years?

Thanks, good point. I have a matrix with all the data that is structured like this (where Y1 is species one, Y2 is species two and so on..., X is site number, Z is the year);

Y1 Y2 X Z

1 1 9 2006

0 1 10 2006

1 1 13 2006

0 0 19 2006

0 1 23 2006

......

0 0 22 2015

I would use some form of indexing to reference different years of data.

So, for example create a new vector for each X? A vector which is a site-dependent index specifying the years the site in question was surveyed.

[Adam Danz]

To get all rows numbers where site number 'j' was surveyed, use find(). For example for site number 9:

rowNums = find(X == 9); 

rowNums will be a vector of row numbers such as [ 3; 22; 31; 48 ...]; If you're looking at row number 22, the previous time site number 9 was surveyed is in row number 3.

[SBinary]

Thank you very much, Adam! Question solved. Much appreciated!

Answer 1

[Peter Perkins]

Joel, it's hard to tell, but it seems very likely that you'd benefit from putting your data in a timetable. "30 per year" sounds like you have timestamps to a resolution smaller than whole years. Timetables let you select data by time.

You could also convert your species and sites to categorical. You might find it more palatable to write

  mydata(mydata.Species1=='Cryptobranchus alleganiensis' & mydata.Site=='Oneonta',:)

to select data.

[SBinary]

Thanks, Peter. Very useful advice! I still have some way to go from writing code that works, to code that is also elegant and efficient. The reason for working in the data format(s) I was in particular case was to conform to the standard used in the ecological model framework/library I am trying to adapt to my needs. I will definitely try to implement your ideas, though, as they would indeed be beneficial.