MATLAB: How to apply a ttest for specific months

statistics

Hello,

I have a large matrix of 2520x192x144, (month x lon x lat). I would like to do a ttest to compare each year with the same month (ttest of year1 Jan, year2 Jan…… year210Jan) for each month (end result = 12 192×144 matrices).

What would be the best way to set this up?

Thank you,

Holly

Best Answer

Hi,

Let the large matrix of dimensions 2520x192x144 be ‘A’. To compare data related to a month for all years, the data is to be separated based on months and t-test is perfomed on each.

month_data=A(month_number:12:2520,:,:); 
h=ttest(month_data);

For Example,

January_data=A(1:12:2520,:,:); 
January_test=ttest(January_data);

This produces a matrix of dimension 1x192x144. Thus all months give 12 matrices.

Related Solutions

MATLAB: How to extract seasonal means

For those just coming here, you can see additional details and explanation here.

You seem to be grouping Jan, Feb and Dec from the same year together. I would be more inclined to group the winters as Dec yr1 with Jan yr2, Feb yr2. In doing so, I see the challenge being.

Grouping Dec from one year with Jan, Feb of the next
The first year will only have Jan, Feb while the last year will only have Dec.
If no data is removed, this means your will have data covereing 38 winters (Jan, Feb in last year as well as Dec in last year).

You asked what you were doing wrong. I don't think cumsum is doing what you want. The first value is itself. The second is the first plus the second. This pattern continues until the last value is the sum of all previous values plus itself. Instead, you want to sum each group separately.

Here's an approach that uses findgroups/splitapply

% Load data from nc file
lon=ncread('sstarc.nc','lon');
lat=ncread('sstarc.nc','lat');
sst=ncread('sstarc.nc','sst');
% Create vector of dates to use for extracting winter months
startDate = datetime(1982,1,1,"Format","MMM-uuuu");
endDate = datetime(2018,12,31,"Format","MMM-uuuu");
dates = startDate:calmonths(1):endDate;
winterMonth = month(dates) == 1 | month(dates) == 2 | month(dates) == 12;
% Extract winter data from sst
sstWinter = sst(:,:,winterMonth);
%% Create grouping variable for winter year 
% (treating consecutive Dec, Jan Feb as winter that year)
% Extract winter year as the year corresponding to January
winterYr = year(dates(winterMonth));
winterYr(month(dates(winterMonth))==12) = winterYr(month(dates(winterMonth))==12)+1;
% Find groups
G = findgroups(winterYr);
% Splitapply can't be used on 3D matrices. Reshaped to 2D to take the mean
sstTemp = reshape(sstWinter,size(sstWinter,1)*size(sstWinter,2),size(sstWinter,3));
f = @(x) mean(x,2,"omitnan");
tempYrMean = splitapply(f,sstTemp,G);
% reshape back to original number of rows, columns
sstYrMean = reshape(tempYrMean,size(sstWinter,1),size(sstWinter,2),[]);
%% Visualizing data
% Create variables for visualizing
[LAT, LON] = meshgrid(lat,lon);
winter1 = sstYrMean(:,:,1); % Change the index of the 3rd dimension to change the year
% Visualize the data for Winter 1982
geoscatter(LAT(:),LON(:),[],winter1(:))
colorbar(gca,"eastoutside")
title("Winter " + num2str(winterYr(1)))

MATLAB: Number of non-nan values across each element in n-dimensional matrix

Is this what you want?

x = your 192x144x8766 array
result = sum(~isnan(x),3);

Best Answer

Related Solutions

MATLAB: How to extract seasonal means

MATLAB: Number of non-nan values across each element in n-dimensional matrix

Related Question