Using Summary on a table to access just the last 2 column stats

Hello, I have an array of numbers that I want to apply the "summary" feature associated with tables to. This is because I want to exploit the stats this gives

The issue Im having is that the number of cloumns and clomn names isnt always the same. what is the same is I really want the stats for the last 2 columns always

This is what I have done so far:

   T=app.UITable;
   d=T.Data;           
   cn=T.ColumnName         
   d2=array2table(d,'VariableNames',cn)   
   %  d2.Properties
   S=summary(d2)   
   % B = string(fieldnames(S))
   % B(end)
   

But is is just a struct, so I wanted to access the last 2 column details which I understand is via

S.fieldname.

But thats my issue, the field name is in cn but doing

   S.cn(end)

doesn't work

Answer 1

[dpb]

You fell into a trap of using a routine that isn't suitable for your purposes...while summary does compute a bunch of statistics without much need for user input, its output wasn't designed to return those parameters for other than viewing.

For your purposes I'd suggest doing something more on the line of

tM=array2table(magic(5));       % out friendly sample dataset yet again...
format bank, format compact     % just to make output consistent...
% our supporting and engine code...
ourstats={'Min','Max','Mean','Median','StD'};     % a list of wanted statistics
fnS=@(x)[min(x);max(x);mean(x);median(x);std(x)]; % define our wanted statistics as anonymous function
Statistics=fnS(tM.(width(tM)-1));                 % compute the statistics as array
tS=array2table(Statistics);                       % convert to table
tS.Properties.RowNames=ourstats                   % add the identifier of which is which

tS = 5x1 table
              Statistics
              __________
    Min          2.00   
    Max         21.00   
    Mean        13.00   
    Median      14.00   
    StD          8.06   

The statistic names can be augmented as desired or the variable identified as the column header, your choice. It's also straightforward to add the units column.

I haven't messed with the uitable enough to know about all the internal formatting it supports although I know there is quite a bit for presentation-only purposes available, but building the base data is pretty straightforward from a "deadahead" path rather than trying to parse a difficult-to-use builtin function output.

Probably you didn't recognize that can build an anonymous function comprising several of the builtin functions outputs catenated as long as they all reduce to a common data type.

NOTA BENE:

Statistics=fnS(tM.(width(tM)-1));

is yet another addressing mode of using the desired column number. width()-1 is the same as the subscripting expression end-1 without needing the array reference for the end function to key against.

[Jason]

Yes i realise now using the table approach and its summary isnt the best.

Thanks forvall your advice

[Jason]

Also, how would you do the concatenation to show the stats for another column:

   tM=array2table(d); 
   ourstats={'Min','Max','Abs Max'};     % a list of wanted statistics
   fnS=@(x)[min(x);max(x);max(abs(x))]; % define our wanted statistics as anonymous function
   dz=fnS(tM.(width(tM)-1));                 % compute the statistics as array instead of end-1
   xtilt=fnS(tM.(width(tM)));  
   tS=array2table([dz,xtilt]);  
   tS.Properties.RowNames=ourstats

didn't work (i.e. didnt give the headings)

                 Var1         Var2   
               _________    _________
    Min        -0.001955    -0.000391
    Max         -1.1e-05     -2.2e-05
    Abs Max     0.001955     0.000391
    

I tried this but no luck

tS.Properties.VariableNames={"dz","xtilt"}

[dpb]

d=randn(5);     % make up some data
tM=array2table(d); 
ourstats={'Min','Max','Abs Max'};     % a list of wanted statistics
fnS=@(x)[min(x);max(x);max(abs(x))]; % define our wanted statistics as anonymous function
dz=fnS(tM.(width(tM)-1));                 % compute the statistics as array instead of end-1
xtilt=fnS(tM.(width(tM)));  
tS=table(dz,xtilt); % individual variables/columns are listed sequentially in table()
tS.Properties.RowNames=ourstats

tS = 3x2 table
                  dz        xtilt 
               _________    ______

    Min          -1.3967    -1.026
    Max        -0.083213    1.2048
    Abs Max       1.3967    1.2048

NOTA BENE, however. To reference more than one column at a time, it's better to revert to the alternate syntax

W=width(tM);                        % save for reuse
stats=fnS(tM{:,W-1:W});             % fn works for arrays, too...
tS=array2table(stats,'VariableNames',{'dz','xtilt'},'RowNames',ourstats)

tS = 3x2 table
                  dz        xtilt 
               _________    ______

    Min          -1.3967    -1.026
    Max        -0.083213    1.2048
    Abs Max       1.3967    1.2048

Of course, if the variable names are known from somewhere else, then can use that instead of hardcoding in the names as above to be more generic code.

[dpb]

"didn't work (i.e. didnt give the headings)"

An array of more than one column will always have variable names Var1, Var2, ... because there are no internal names associated with individual columns for array2table to use. table with a list of variables has those available to it so the behavior is different in that regards. You can assign variable names with the named parameter as above; if the variable names are known from somewhere else, then can use that instead to be more generic code.

What happened in your assignment by using the Properties.VariableNames is not possible to diagnose without the complete session; it is valid syntax and will work if something else we can't see here didn't go wrong. An accompanying error message would prossibly identify what that something was.

[Jason]

Thankyou for all your excellent explanations

[dpb]

Glad to help...the table(*) and all its possible addressing rules can get pretty complex; it takes some "time in grade" and a lot of study/practice to learn one's way around thoroughly.

(*) As is also the struct

Answer 2

[Steven Lord]

As shown on this documentation page, one of the ways to extract a sub-table from a table is numeric indices.

M = magic(5);
sampleTable = array2table(M)

sampleTable = 5x5 table
    M1    M2    M3    M4    M5
    __    __    __    __    __

    17    24     1     8    15
    23     5     7    14    16
     4     6    13    20    22
    10    12    19    21     3
    11    18    25     2     9

I could use variable names:

subtable1 = sampleTable(:, ["M4", "M5"])

subtable1 = 5x2 table
    M4    M5
    __    __

     8    15
    14    16
    20    22
    21     3
     2     9

Or I could use numbers.

subtable2 = sampleTable(:, [4 5])

subtable2 = 5x2 table
    M4    M5
    __    __

     8    15
    14    16
    20    22
    21     3
     2     9

Using end in the indexing expression also works.

subtable3 = sampleTable(:, [end-1 end])

subtable3 = 5x2 table
    M4    M5
    __    __

     8    15
    14    16
    20    22
    21     3
     2     9

[dpb]

Another syntax way that uses variable names with the [end-1:end] indexing

M = magic(5);
sampleTable = array2table(M)

sampleTable = 5x5 table
    M1    M2    M3    M4    M5
    __    __    __    __    __

    17    24     1     8    15
    23     5     7    14    16
     4     6    13    20    22
    10    12    19    21     3
    11    18    25     2     9

sampleTable(:,sampleTable.Properties.VariableNames([end-1:end]))

ans = 5x2 table
    M4    M5
    __    __

     8    15
    14    16
    20    22
    21     3
     2     9

This syntax is useful if one knows a given matching name or pattern of names but the position isn't necessarily fixed.

[Steven Lord]

Yes, you could match patterns or use string processing functions if you operate on the variable names.

M = magic(5);
sampleTable = array2table(M, ...
    VariableNames = ["ace", "king", "queen", "jack", "ten"])

sampleTable = 5x5 table
    ace    king    queen    jack    ten
    ___    ____    _____    ____    ___

    17      24       1        8     15 
    23       5       7       14     16 
     4       6      13       20     22 
    10      12      19       21      3 
    11      18      25        2      9 

namesContainA = sampleTable(:, ...
    contains(sampleTable.Properties.VariableNames, 'a'))

namesContainA = 5x2 table
    ace    jack
    ___    ____

    17       8 
    23      14 
     4      20 
    10      21 
    11       2 

[Jason]

Thanks Steven.

but i need to get the stats and for that i need s.fieldname.

I don't know in advance what the column headings are (= the field names), but I can get it via

T=app.UITable3;
cn=T.ColumnName

So is there a way to use the S.fieldname by just getting the last field name from cn, somehting like

S=summary(d2)

stats_I_want=S.cn{end}

[Steven Lord]

You could use indexing or dynamic {field, variable, property} names, depending on whether S is a {struct, table, object}.

M = magic(5);
sampleTable = array2table(M)

sampleTable = 5x5 table
    M1    M2    M3    M4    M5
    __    __    __    __    __

    17    24     1     8    15
    23     5     7    14    16
     4     6    13    20    22
    10    12    19    21     3
    11    18    25     2     9

n = randi(width(M), 1)

n = 5

varname = "M"+n

varname = "M5"

S1 = sampleTable.(varname) % double vector

S1 = 5×1

15 16 22 3 9

<mw-icon class=""></mw-icon>

<mw-icon class=""></mw-icon>

●

S2 = sampleTable(:, varname) % table

S2 = 5x1 table
    M5
    __

    15
    16
    22
     3
     9

S3 = sampleTable{:, varname} % double vector

S3 = 5×1

15 16 22 3 9

<mw-icon class=""></mw-icon>

<mw-icon class=""></mw-icon>

Answer 3

[dpb]

Use the previous illustrated syntax on the summary table, too...you said you knew the columns of interest are the last two, so

tM=array2table(magic(5));
S=summary(tM(:, [end-1 end]))

S = struct with fields:
    Var4: [1x1 struct]
    Var5: [1x1 struct]

S.Var4

ans = struct with fields:
           Size: [5 1]
           Type: 'double'
    Description: ''
          Units: ''
     Continuity: []
     NumMissing: 0
            Min: 2
         Median: 14
            Max: 21
           Mean: 13
            Std: 8.0623

fnames=fieldnames(S)

fnames = 2x1 cell array
    {'Var4'}
    {'Var5'}

stats_I_want=S.(char(fnames(end)))

stats_I_want = struct with fields:
           Size: [5 1]
           Type: 'double'
    Description: ''
          Units: ''
     Continuity: []
     NumMissing: 0
            Min: 3
         Median: 15
            Max: 22
           Mean: 13
            Std: 7.2457

It might be more simpler to use groupsummary:here depending on just what you are looking for even though may have to create a fake grouping variable...

tM1=tM(:, [end-1 end]);
tM1=addvars(tM1,ones(height(tM1),1),'NewVariableNames','Group','Before',tM1.Properties.VariableNames(1))

tM1 = 5x3 table
    Group    Var4    Var5
    _____    ____    ____

      1        8      15 
      1       14      16 
      1       20      22 
      1       21       3 
      1        2       9 

S=groupsummary(tM1,'Group','all',tM1.Properties.VariableNames(end))

S = 1x14 table
    Group    GroupCount    mean_Var5    sum_Var5    min_Var5    max_Var5    range_Var5    median_Var5    mode_Var5    var_Var5    std_Var5    nummissing_Var5    nnz_Var5    numunique_Var5
    _____    __________    _________    ________    ________    ________    __________    ___________    _________    ________    ________    _______________    ________    ______________

      1          5            13           65          3           22           19            15             3          52.5       7.2457            0              5              5       

You can select any subset of the above and/or add others as well...

Alternatively, with it you can select which variables to use as the data variables as input rather than subselecting another table...

datavars=false(1,width(tM)); datavars(end)=true;   % select only last variable
tM=addvars(tM,ones(height(tM),1),'NewVariableNames','Group','Before',1)

tM = 5x6 table
    Group    Var1    Var2    Var3    Var4    Var5
    _____    ____    ____    ____    ____    ____

      1       17      24       1       8      15 
      1       23       5       7      14      16 
      1        4       6      13      20      22 
      1       10      12      19      21       3 
      1       11      18      25       2       9 

S=groupsummary(tM,'Group','all',datavars)

S = 1x14 table
    Group    GroupCount    mean_Var4    sum_Var4    min_Var4    max_Var4    range_Var4    median_Var4    mode_Var4    var_Var4    std_Var4    nummissing_Var4    nnz_Var4    numunique_Var4
    _____    __________    _________    ________    ________    ________    __________    ___________    _________    ________    ________    _______________    ________    ______________

      1          5            13           65          2           21           19            14             2           65        8.0623            0              5              5       

[Jason]

Thankyou, this is exactly what I was looking for to get the last fieldname (=column heading in uitable) when it could be anything. I'd also like to appreciate Stevens effort too which has helped me.

stats_I_want=S.(char(fnames(end)))

Could I ask one more request.

I'd like to display (for the last two columns only), a table in a uitextarea showing the stats

I use this:

   % Table approach
   T=app.UITable3;
   d=T.Data; 
   cn=T.ColumnName;          
   d2=array2table(d,'VariableNames',cn);   % ,'VariableNames',{'Feet','Inches','Centimeters'}))
   S=summary(d2)
   fnames=fieldnames(S) 
   S2=S.(char(fnames(end)))
   S3=S.(char(fnames(end-1)))
   str = formattedDisplayText(S2)   %My add to message area function, see below 
   

where

   function ReportMessage(app,msg)
        currString=get(app.MessagesTextArea,'Value');
        %currString=[{char(msg)};currString];  %add to top of message box
        currString=[currString; {char(msg)}];  %add to bottom of message box
        app.MessagesTextArea.Value=currString;
        drawnow;
        scroll(app.MessagesTextArea,'bottom');
    end

This displays this in my textarea

            Size: [667 1]
           Type: 'double'
    Description: ''
          Units: ''
     Continuity: []
     NumMissing: 0
            Min: -0.000391
         Median: -0.000235
            Max: -2.2e-05
           Mean: -0.00021689
            Std: 0.00013186 

1: How can I just show these parameters, and justify it like this:

 % column heading #1 (=end)        column heading #2 (=end-1)    (underlined)
 Min:     -0.000391                % values for this column
 Median:  -0.000235                % values for this column
 Max:     -2.2e-05                 % values for this column
 Mean:    -0.00021689              % values for this column
 Std:     0.00013186               % values for this column

This is why I used the "table approach" because i thought I could just splat the ouptut of the table summary into the uiarea with a "table" appearance

[Jason]

This seems to work:

T=app.UITable;
d=T.Data; 
Mx=max(d);   % Gets max of all columns by default
Mn=min(d);   % Gets min of all columns by default

% Create Table manually
params = ["Max_Xtilt";"Min_Xtilt";"Max_dx(um)";"Min_dz(um)"];
Vals = [10^6*Mx(1,end);10^6*Mn(1,end);10^6*Mx(1,end-1);10^6*Mn(1,end-1)];
units = ["urads";"urads";"um";"um"];

tbl = table(params,Vals,units)
str = formattedDisplayText(tbl)
str = regexprep(str,'</?strong>','')
ReportMessage(app,str)

     params       Vals      units 
    ____________    _____    _______
    "Max_Xtilt"       -22    "urads"
    "Min_Xtilt"      -391    "urads"
    "Max_dx(um)"      -11    "um"   
    "Min_dz(um)"    -1955    "um"