Thread: what is the best statistical test for this problem?

1. what is the best statistical test for this problem?

i have a set of a few tens of satellite images that show the sea surface temperature (SST) for a span of ocean surface measuring about 100km by 100km. pixel size is about 500m by 500m. these images are taken over the same area, over a time period of about 10 years. i want to find out from these images if there have been any statistically-significant changes in the SST over this time period. any idea what statistical test i can use?

using something like an average value for each image would not be ideal, as a lot of information would be lost by condensing thousands of numbers to just 1. on the other hand, i can divide each image into several smaller segments, work out the average for each segment, and compare these averages over time. the trick here would then be to find segments which are representative of the entire image, and these segment choices might not be convincing. any ideas?

2. Have you plotted these measurements in a time series to look for any trends?

Before you think about which test to use you should ask further questions like are the samples all taken around the same time of year? Otherwise you could have a bias sample. I like the idea of using the smaller segments, the trick with these are to sample the same segment each time and compare those.

3. nope, i have yet to try plotting them out. actually, what i should be asking is, what approach is appropriate in this case to determine if there has been any change in the SST?

4. I would plot the data first.

Otherwise you may look at a test for difference in means given there is no bias and you know the distribution of the data.