Simbarashe, readtable and readmatrix read your file just fine, so I don't think that's the issue:
>> sst2 = readmatrix("sst2.txt");
Name Size Bytes Class Attributes
sst2 180x360 518400 double
You have not provided the lat and lon vectors, and I'm not really able to follow your code to glean what they are. But if what you want is the data in a rectangular subset of your data (that assumes it's on a rectangular grid in lat/lon space), I would think you can just use row/column indexing and skip the thing with meshgrid and inpolygon. Something like
>> nanmean(sst2(1:10,1:10))
-1.686 -1.6809 -1.6667 -1.6505 -1.6336 -1.6096 -1.574 -1.5394 -1.5111 -1.4503
>> nanmean(sst2(1:10,1:10),"all")
But it's pretty hard to tell what you have. If the output of nanmean is really [] as you say, then you have an issue with idx, which, as I said, seems like you can avoid that whole calculation.
