Creating M histograms from an NxM table on separate plots

Hi there,

I have an N x M table and I am trying to create M histograms on different plots, I can't seem to figure out how to plot them on different figures. If I hold on, they all plot on the same axis'

data = readtable('file.csv'); % Import in wine raw data
predictors = data(:,1:width(data)-1); % remove quality as it is our output
headers = predictors.Properties.VariableNames; % Get predictor labels
%% Lets check to see how our predictors are distributed
statarray = grpstats(predictors,[],{'mean','std'})
for k = width(headers);
    a = table2array(predictors(:,k)); % Converting from table to array for hist
    histogram(a)
    title(headers(1,k))
end

I'm also not sure why I can't pass a table row to a histogram and need to table2array the data.

Any help would be greatly appreciated. Thank you

[Jeremy Brecevic]

Huge. Thank you mate

Answer 1

[Steven Lord]

A = 2*randn(1e4, 6) + 5*(0:5);
T = array2table(A);
for k = 1:width(T)
    subplot(2, 3, k)
    histogram(T{:, k})
end

Note that each histogram is roughly centered around a different multiple of five, but each has (roughly) the same shape.

Answer 2

[dpb]

...
for k=1:width(headers)              % k=width(headers) --> one value, the last column
  figure                            % create a new figure for each histogram plot
  histogram(predictors.(predictors.Properties.VariableNames(k)); % hist each variable
  title(predictors.Properties.VariableNames{k})
end

using your secondary table. There's no reason to duplicate data, however, just use the variables of interest in the original table:

data = readtable('file.csv'); % Import in wine raw data
statarray=grpstats(data,[],{'mean','std'},'DataVars'data.Properties.VariableNames(1:width(data)-1));
for k=1:width(1:width(data)-1))
  figure                                                % create a new figure for each histogram plot
  histogram(data.(data.Properties.VariableNames{k});    % hist each variable
  title(data.Properties.VariableNames{k})
end

"not sure why I can't pass a table row to a histogram"

It's not a row but a column...but, that aside, you referenced the table with () subscripting which returns another table. To dereference the table use {} subscripting similar to a cell array. Or use "dot" addressing with a table variable name as illustrated above. There is a veritable cornucopia of addressing modes possible with tables; see <Access-data-in-a-table> for the details.

[Jeremy Brecevic]

Hi, Thanks for this, unfortunately I run into the same issue as before I table2array'd the data where I get a type mismatch when trying to use dots notation

Invalid expression. When calling a function or indexing a variable, use parentheses. Otherwise, check for mismatched delimiters.

Using Figure, I also only get the last chart to display.

[dpb]

Show your code...and error in context. The above worked here with a table.

You'll get as many figures as times through the loop...if it aborts early, then it's obvious you won't get them all. And, of course, each figure is opened on top of (or nearly so) so you may only see the one even though they're all there.

Steven Lord's idea of subplot may be what you were thinking of as desired format -- I took the request as stated literally as one/figure. With a little table that happened to be here already--

>> t
t =
  10×5 table
    Time    P1       P2        V         S   
    ____    ___    ______    ______    ______
    0.01    100    14.142    29.022    5.3872
    0.02    110    14.832    30.555    5.5276
    0.03    120    15.492    32.019    5.6585
    0.04    130    16.125    33.423    5.7813
    0.05    140    16.733    34.774     5.897
    0.06    150    17.321    36.078    6.0065
    0.07    160    17.889    37.338    6.1105
    0.08    170    18.439    38.559    6.2096
    0.09    180    18.974    39.745    6.3044
     0.1    190    19.494    40.898    6.3952
>> figure
>> j=0;
>> for i=2:width(t)
j=j+1;
subplot(2,2,j)
histogram(t.(t.Properties.VariableNames{j}))
end
>> 

resulted in the following:

Not all that interesting, but proves the code as written will work if you don't make some other syntax error.

Steven's code illustrated the alternative of {} dereferencing by column number rather than variable name.

But, the key point is, you MUST use a form that returns the dereference data from the table, not another table. As the referenced link to the documentation shows, there are many ways to address a table; many of them return another table; that appears to be the error you're still making.