Printing output of fitglm as table to file

Dear all,

I have fitted my regression with fitglm, inputting a table. Therefore, the output looks like a nice table with named regressors etc.

What is the best way to save this as a file for publication? An html table would be good or I guess a figure would be ok too.

Many thanks

Jacquie

[Adam Danz]

Why not just print it out to text file rather than printing it to a figure or html?

[Jacqueline Scholl]

that would work. how do I do that please?

[Adam Danz]

The answer below should get you started. It prints out a lot of the tables and other data from the mdl output. It's a complete working example so you can step through it to produce the data and the text file. Then you can adapt it to your needs.

[Jacqueline Scholl]

wow, thank you so much! that would have taken me forever to figure out!

[Adam Danz]

Glad I could help!

Answer 1

[Adam Danz]

Update: MATLAB R2021a or later

Use the formattedDisplayText function to capture the model output as a string (see Community Highlight).

%% Create a demo model
load hospital %built-in matlab data
dsa = hospital;
modelspec = 'Smoker ~ Age*Weight*Sex - Age:Weight:Sex';
mdl = fitglm(dsa,modelspec,'Distribution','binomial');

% Convert summary to char array
txt = formattedDisplayText(mdl)

Method 1: Convert model summary to char array

%% Create a demo model
load hospital %built-in matlab data
dsa = hospital;
modelspec = 'Smoker ~ Age*Weight*Sex - Age:Weight:Sex';
mdl = fitglm(dsa,modelspec,'Distribution','binomial');

% Convert summary to char array
txt = evalc('mdl')

Method 2: Extract parts of model

This method extracts each section of text from the model and stores the text in a nx1 cell array named txt. The cell aray of text will then be written to a text file.

%% Create a demo model
load hospital %built-in matlab data
dsa = hospital;
modelspec = 'Smoker ~ Age*Weight*Sex - Age:Weight:Sex';
mdl = fitglm(dsa,modelspec,'Distribution','binomial');



%% Create text and tables
% Create rows of text that should come before the table 
txt{1,1} = mdl.Formula.char;
txt{2,1} = sprintf('Distribution = %s',mdl.Distribution.Name);

% Create the table of coefficients.  I also appended the confidence intervals to the end.
txt{3,1} = table2CellStrFcn([mdl.Coefficients, array2table(mdl.coefCI,'VariableNames',{'coefCIlow','coefCIhigh'})]); 

% Create rows of text that should come after the table 
txt{4,1} = sprintf('%d observations, %d df', mdl.NumObservations, mdl.DFE);
[pv,fstat] = coefTest(mdl);
txt{5,1} = sprintf('F-Statistic vs. constant model: %.1f, p = %0.6f', fstat, pv);

% Create additional tables if needed
struct2str = @(x)strsplit(evalc('disp(x);'),newline)'; 
txt{6,1} = ['________R-Squared________'; struct2str(mdl.Rsquared)];
txt{7,1} = ['_____Criterion_____'; struct2str(mdl.ModelCriterion)];
varinfoTable = mdl.VariableInfo;
varinfoTable.Class = []; 
varinfoTable.Range = [];
txt{8,1} = table2CellStrFcn(varinfoTable); 



%% Write text and tables to file. 
% Convert any non-cell elements of txt to cell
notCellIdx = ~cellfun(@iscell, txt); 
txt(notCellIdx) = cellfun(@(x){{x}},txt(notCellIdx)); 
% add an empty row between each element to separate the sections of text
txtSpace = repmat({{' '}}, size(txt)); 
txt = reshape([txt';txtSpace'],[],1); 
% Vertically concatenate cell arrays
txt = vertcat(txt{:}); 

% Open / Create a new text file named 'GLM_results.txt'. 
fid = fopen('GLM_results.txt', 'wt'); % a full path would be better
fprintf(fid,'%s\n', txt{:}); 
fclose(fid); 



%% Function that converts table to cell of strings
function Tstr = table2CellStrFcn(T)
% Input a numeric table with column and row names (T) and convert to cell of strings (Tstr)
cname = [' ', T.Properties.VariableNames]; 
rname = T.Properties.RowNames; 
datastr = cellfun(@num2str,table2cell(T),'UniformOutput',false); 
Tstr = [cname; [rname, datastr]]; 
% pad elements of Tstr so columns have the same length
maxLen = max(cellfun(@numel, Tstr),[], 1); 
for i = 1:size(Tstr,2)
    Tstr(:,i) = pad(Tstr(:,i),maxLen(i),'right'); 
end
% Join columns to make nx1 array
Tstr = cellfun(@strjoin,mat2cell(Tstr,ones(1,size(Tstr,1)), size(Tstr,2)),'Unif',false); 
end

(Updated on 6/21/22 to define struct2str)

Here's an screen shot of the resultant text file.

Answer 2

[Vlad Pineta]

I've created the same file but exporting as LaTex and without the error about "struct2str"

function regressionExport(name,mdl,vargin)
txt{1,1} = '\begin{table}[]';
txt{2,1} = '\centering';
txt{3,1} = '\begin{tabular}{|ccccccc}';
txt{4,1} = '\hline';
txt{5,1} = strcat('\multicolumn{7}{|c|}{',mdl.Formula.char,'}',{' '},'\\ \hline');
txt{6,1} = strcat('\multicolumn{7}{|c|}{','Distribution =',{' '},mdl.Distribution.Name,'}',' \\ \hline');
tabel_coefficients = [mdl.Coefficients, array2table(mdl.coefCI,'VariableNames',{'coefCIlow','coefCIhigh'})];
tabel_coefficients_cname = [' ', tabel_coefficients.Properties.VariableNames];
tabel_coefficients_rname = tabel_coefficients.Properties.RowNames;
tabel_coefficients_datastr = cellfun(@num2str,table2cell(tabel_coefficients),'UniformOutput',false);
tabel_coefficients = [tabel_coefficients_cname; [tabel_coefficients_rname, tabel_coefficients_datastr]];
tabel_coefficients = string(tabel_coefficients);
txt{7,1} = strcat(tabel_coefficients{1,1},{' '},'&',{' '},tabel_coefficients{1,2},{' '},'&',{' '},tabel_coefficients{1,3},{' '},'&',{' '},tabel_coefficients{1,4},{' '},'&',{' '},tabel_coefficients{1,5},{' '},'&',{' '},tabel_coefficients{1,6},{' '},'& \multicolumn{1}{c|}{',tabel_coefficients{1,7},'} \\ \cline{2-7}');
tabel_coefficients(1,:) = [];
txt8 = {};
for i = 1:size(tabel_coefficients,1)
    if i == size(tabel_coefficients,1)
        x = strcat('\multicolumn{1}{|c|}{',tabel_coefficients(i,1),'} &',{' '},tabel_coefficients(i,2),{' '},'&',{' '},tabel_coefficients(i,3),{' '},'&',{' '},tabel_coefficients(i,4),{' '},'&',{' '},tabel_coefficients(i,5),{' '},'&',{' '},tabel_coefficients(i,6),{' '},'& \multicolumn{1}{c|}{',tabel_coefficients(i,7),'} \\ \hline');
        txt8 = [txt8;x];
    else
        x = strcat('\multicolumn{1}{|c|}{',tabel_coefficients(i,1),'} &',{' '},tabel_coefficients(i,2),{' '},'&',{' '},tabel_coefficients(i,3),{' '},'&',{' '},tabel_coefficients(i,4),{' '},'&',{' '},tabel_coefficients(i,5),{' '},'&',{' '},tabel_coefficients(i,6),{' '},'& \multicolumn{1}{c|}{',tabel_coefficients(i,7),'} \\');
        txt8 = [txt8;x];
    end
end
txt{8,1} = txt8;
txt{9,1} = strcat('\multicolumn{7}{|c|}{',string(mdl.NumObservations),' observations ,',{' '},string(mdl.DFE),' df','} \\ \hline');
[pv,fstat] = coefTest(mdl);
txt{10,1} = strcat('\multicolumn{7}{|c|}{',sprintf('F-Statistic vs. constant model: %.1f, p = %0.6f', fstat, pv),'} \\ \hline');
txt{11,1} = strcat('\multicolumn{2}{|c|}{R-Squared} &  &  &  &  &  \\ \cline{1-2}');
tabel_rsquared = struct2table(mdl.Rsquared);
cell_rsquared = table2cell(tabel_rsquared);
titles_rsquared = tabel_rsquared.Properties.VariableNames;
cell_rsquared = [titles_rsquared;cell_rsquared];
cell_rsquared = cell_rsquared';
cell_rsquared = string(cell_rsquared);
txt12 = {};
for i = 1:size(cell_rsquared,1)
    if i == size(cell_rsquared,1)
        x = strcat(cell_rsquared(i,1),{' '},'& \multicolumn{1}{c|}{',cell_rsquared(i,2),'} &  &  &  &  &  \\ \cline{1-2}');
        txt12 = [txt12;x];
    else
        x = strcat(cell_rsquared(i,1),{' '},'& \multicolumn{1}{c|}{',cell_rsquared(i,2),'} &  &  &  &  &  \\');
        txt12 = [txt12;x];
    end
end
txt{12,1} = txt12;
txt{13,1} = strcat('\multicolumn{2}{|c|}{Criterion} &  &  &  &  &  \\ \cline{1-2}');
tabel_modelcriterion = struct2table(mdl.ModelCriterion);
cell_modelcriterion = table2cell(tabel_modelcriterion);
titles_modelcriterion = tabel_modelcriterion.Properties.VariableNames;
cell_modelcriterion = [titles_modelcriterion;cell_modelcriterion];
cell_modelcriterion = cell_modelcriterion';
cell_modelcriterion = string(cell_modelcriterion);
txt14 = {};
for i = 1:size(cell_modelcriterion,1)
    if i == size(cell_modelcriterion,1)
        x = strcat(cell_modelcriterion(i,1),{' '},'& \multicolumn{1}{c|}{',cell_modelcriterion(i,2),'} &  &  &  &  &  \\ \cline{1-3}');
        txt14 = [txt14;x];
    else
        x = strcat(cell_modelcriterion(i,1),{' '},'& \multicolumn{1}{c|}{',cell_modelcriterion(i,2),'} &  &  &  &  &  \\');
        txt14 = [txt14;x];
    end
end
txt{14,1} = txt14;
tabel_variableinfo = mdl.VariableInfo;
tabel_variableinfo.Class = [];
tabel_variableinfo.Range = [];
tabel_variableinfo_cname = [' ', tabel_variableinfo.Properties.VariableNames];
tabel_variableinfo_rname = tabel_variableinfo.Properties.RowNames;
tabel_variableinfo_datastr = cellfun(@num2str,table2cell(tabel_variableinfo),'UniformOutput',false);
tabel_variableinfo = [tabel_variableinfo_cname; [tabel_variableinfo_rname, tabel_variableinfo_datastr]];
tabel_variableinfo = string(tabel_variableinfo);
txt{15,1} = strcat(tabel_variableinfo{1,1},{' '},'&',{' '},tabel_variableinfo{1,2},{' '},'& \multicolumn{1}{c|}{',tabel_variableinfo{1,3},'} &  &  & \multicolumn{1}{l}{} &  \\ \cline{2-3}');
tabel_variableinfo(1,:) = [];
txt16 = {};
for i = 1:size(tabel_variableinfo,1)
    if i == size(tabel_variableinfo,1)
        x = strcat('\multicolumn{1}{|c|}{',tabel_variableinfo(i,1),'} &',{' '},tabel_variableinfo(i,2),{' '},'& \multicolumn{1}{c|}{',tabel_variableinfo(i,3),'} &  & \multicolumn{1}{l}{} & \multicolumn{1}{l}{} &  \\ \cline{1-3}');
        txt16 = [txt16;x];
    else
        x = strcat('\multicolumn{1}{|c|}{',tabel_variableinfo(i,1),'} &',{' '},tabel_variableinfo(i,2),{' '},'& \multicolumn{1}{c|}{',tabel_variableinfo(i,3),'} &  & \multicolumn{1}{l}{} & \multicolumn{1}{l}{} &  \\');
        txt16 = [txt16;x];
    end
end
txt{16,1} = txt16;
txt{17,1} = '\end{tabular}';
txt{18,1} = '\end{table}';
notCellIdx = ~cellfun(@iscell, txt);
txt(notCellIdx) = cellfun(@(x){{x}},txt(notCellIdx));
txt = vertcat(txt{:});
fid = fopen(strcat(name,'.txt'), 'wt');
fprintf(fid,'%s\n', txt{:});
fclose(fid);
end

I know it's not the best code but it works :)

[Adam Danz]

Thanks for pointing out the struct2str error. I fixed that in my answer.

[Vlad Pineta]

I've modified the code to be able to print the results as pdf and i removed the inModel and isCategorical fields for the covariance coefficient. Also i rearranged the R-squared and Criterion fields for a nicer view.

function export = fitlmExport(rEdataFolder,rEname,mdl,vargin)
optionExport = 1;
if optionExport == 1
    txt{1,1} = {'\documentclass[convert]{standalone}';
        '\begin{document}';
        '\begin{minipage}{1.6\textwidth}'};
end
txt{2,1} = '\begin{table}[!htb]';
txt{3,1} = '\centering';
txt{4,1} = '\begin{tabular}{|ccccccc|}';
txt{5,1} = '\hline';
txt{6,1} = strcat('\multicolumn{7}{|c|}{',rEname,'}',{' '},'\\ \hline');
txt{7,1} = strcat('\multicolumn{7}{|c|}{',mdl.Formula.char,'}',{' '},'\\ \hline');
tabel_coefficients = [mdl.Coefficients, array2table(mdl.coefCI,'VariableNames',{'coefCIlow','coefCIhigh'})];
tabel_coefficients_cname = [' ', tabel_coefficients.Properties.VariableNames];
tabel_coefficients_rname = tabel_coefficients.Properties.RowNames;
tabel_coefficients_datastr = cellfun(@num2str,table2cell(tabel_coefficients),'UniformOutput',false);
tabel_coefficients = [tabel_coefficients_cname; [tabel_coefficients_rname, tabel_coefficients_datastr]];
tabel_coefficients = string(tabel_coefficients);
txt{8,1} = strcat('\multicolumn{1}{|c}{}',tabel_coefficients{1,1},{' '},'&',{' '},tabel_coefficients{1,2},{' '},'&',{' '},tabel_coefficients{1,3},{' '},'&',{' '},tabel_coefficients{1,4},{' '},'&',{' '},tabel_coefficients{1,5},{' '},'&',{' '},tabel_coefficients{1,6},{' '},'&',{' '},tabel_coefficients{1,7},{' '},'\\ \cline{2-7}');
tabel_coefficients(1,:) = [];
txt9 = {};
for i = 1:size(tabel_coefficients,1)
    if i == size(tabel_coefficients,1)
        x = strcat('\multicolumn{1}{|c|}{',tabel_coefficients(i,1),'} &',{' '},tabel_coefficients(i,2),{' '},'&',{' '},tabel_coefficients(i,3),{' '},'&',{' '},tabel_coefficients(i,4),{' '},'&',{' '},tabel_coefficients(i,5),{' '},'&',{' '},tabel_coefficients(i,6),{' '},'&',{' '},tabel_coefficients(i,7),{' '},'\\ \hline');
        txt9 = [txt9;x];
    else
        x = strcat('\multicolumn{1}{|c|}{',tabel_coefficients(i,1),'} &',{' '},tabel_coefficients(i,2),{' '},'&',{' '},tabel_coefficients(i,3),{' '},'&',{' '},tabel_coefficients(i,4),{' '},'&',{' '},tabel_coefficients(i,5),{' '},'&',{' '},tabel_coefficients(i,6),{' '},'&',{' '},tabel_coefficients(i,7),{' '},'\\');
        txt9 = [txt9;x];
    end
end
txt{9,1} = txt9;
txt{10,1} = strcat('\multicolumn{7}{|c|}{',string(mdl.NumObservations),' observations ,',{' '},string(mdl.DFE),' df','} \\ \hline');
[pv,fstat] = coefTest(mdl);
txt{11,1} = strcat('\multicolumn{7}{|c|}{',sprintf('F-Statistic vs. constant model: %.1f, p = %0.6f', fstat, pv),'} \\ \hline');
txt{12,1} = '\multicolumn{7}{l}{} \\ \cline{2-7} ';
tabel_rsquared = struct2table(mdl.Rsquared);
cell_rsquared = table2cell(tabel_rsquared);
titles_rsquared = tabel_rsquared.Properties.VariableNames;
cell_rsquared = [titles_rsquared;cell_rsquared];
cell26 = repmat({''}, 2, 6);
cell_rsquared = [cell_rsquared,cell26];
cell_rsquared = string(cell_rsquared);
txt{13,1} = strcat('\multicolumn{1}{c|}{} &',{' '},cell_rsquared(1,1),{' '},'&',{' '},cell_rsquared(1,2),{' '},'&',{' '},cell_rsquared(1,3),{' '},'&',{' '}',cell_rsquared(1,4),{' '},'&',{' '},cell_rsquared(1,5),{' '},'&',{' '},cell_rsquared(1,6),{' '},'\\ \cline{1-1}');
txt{14,1} = strcat('\multicolumn{1}{|c}{R-Squared} &',{' '},cell_rsquared(2,1),{' '},'&',{' '},cell_rsquared(2,2),{' '},'&',{' '},cell_rsquared(2,3),{' '},'&',{' '},cell_rsquared(2,4),{' '},'&',{' '},cell_rsquared(2,5),{' '},'&',{' '},cell_rsquared(2,6),{' '},'\\ \hline');
tabel_modelcriterion = struct2table(mdl.ModelCriterion);
cell_modelcriterion = table2cell(tabel_modelcriterion);
titles_modelcriterion = tabel_modelcriterion.Properties.VariableNames;
cell_modelcriterion = [titles_modelcriterion;cell_modelcriterion];
cell26 = repmat({''}, 2, 6);
cell_modelcriterion = [cell_modelcriterion,cell26];
cell_modelcriterion = string(cell_modelcriterion);
txt{15,1} = strcat('\multicolumn{1}{c|}{} &',cell_modelcriterion(1,1),{' '},'&',{' '},cell_modelcriterion(1,2),{' '},'&',{' '},cell_modelcriterion(1,3),{' '},'&',{' '},cell_modelcriterion(1,4),{' '},'&',{' '},cell_modelcriterion(1,5),{' '},'&',{' '},cell_modelcriterion(1,6),{' '},'\\ \cline{1-1}');
txt{16,1} = strcat('\multicolumn{1}{|c}{Criterion} &',{' '},cell_modelcriterion(2,1),{' '},'&',{' '},cell_modelcriterion(2,2),{' '},'&',{' '},cell_modelcriterion(2,3),{' '},'&',{' '},cell_modelcriterion(2,4),{' '},'&',{' '},cell_modelcriterion(2,5),{' '},'&',{' '},cell_modelcriterion(2,6),{' '},'\\ \hline');
txt{17,1} = '\end{tabular}';
txt{18,1} = '\end{table}';
cell_CoefficientCovariance = num2cell(mdl.CoefficientCovariance);
cell_CoefficientNames = mdl.CoefficientNames;
cell_CoefficientCovariance = [cell_CoefficientNames;cell_CoefficientCovariance];
cell_CoefficientNames = ['Coefficient Covariance',cell_CoefficientNames];
cell_CoefficientNames = cell_CoefficientNames';
cell_CoefficientCovariance = [cell_CoefficientNames,cell_CoefficientCovariance];
cell_CoefficientCovariance = string(cell_CoefficientCovariance);
txt{19,1} = '\begin{table}[]';
txt{20,1} = '\centering';
txt{21,1} = strcat('\begin{tabular}{|',repelem('c',size(cell_CoefficientCovariance,1)),'|}');
txt{22,1} = '\hline';
txt23 = {};
txt23_row = {};
for j = 1:size(cell_CoefficientCovariance,1)
    for i = 1:size(cell_CoefficientCovariance,2)
        if i == 1 && j == 1
            temp_txt23 = strcat({' '},cell_CoefficientCovariance(j,i),{' '},'&');
        elseif i == size(cell_CoefficientCovariance,2) && j == 1
            temp_txt23 = strcat({' '},cell_CoefficientCovariance(j,i),{' '},'\\ \cline{2-',string(size(cell_CoefficientCovariance,2)),'}');
        elseif i == 1 && j ~= 1
            temp_txt23 = strcat('\multicolumn{1}{|c|}{',cell_CoefficientCovariance(j,i),'}',{' '},'&');
        elseif i == size(cell_CoefficientCovariance,2) && j ~= size(cell_CoefficientCovariance,1)
            temp_txt23 = strcat({' '},cell_CoefficientCovariance(j,i),{' '},'\\');
        elseif i == size(cell_CoefficientCovariance,2) && j == size(cell_CoefficientCovariance,1)
            temp_txt23 = strcat({' '},cell_CoefficientCovariance(j,i),{' '},'\\ \hline');
        else
            temp_txt23 = strcat({' '},cell_CoefficientCovariance(j,i),{' '},'&');
        end
        txt23_row = [txt23_row,temp_txt23];
    end
    txt23 = [txt23;txt23_row];
    txt23_row = {};
end
txt23=cellfun(@string,txt23);
txt{23,1} = join(txt23);
txt{24,1} = '\end{tabular}';
txt{25,1} = '\end{table}';
if optionExport == 1
    txt{26,1} = {'\end{minipage}';
        '\end{document}'};
end
for i = 1:size(txt,1)
    txt{i,1} = regexprep(txt{i,1},'([$_^])','\.');
end
notCellIdx = ~cellfun(@iscell, txt);
txt(notCellIdx) = cellfun(@(x){{x}},txt(notCellIdx));
txt = vertcat(txt{:});
export = sprintf('%s\n', txt{:});
if optionExport == 1
    fid = fopen(strcat(rEdataFolder,rEname,'.txt'), 'wt');
    fprintf(fid,'%s\n', txt{:});
    fclose(fid);
    fid = fopen(strcat(rEdataFolder,rEname,'.tex'), 'wt');
    fprintf(fid,'%s\n', txt{:});
    fclose(fid);
    cmd = ['cd ' rEdataFolder ' && latex ' rEname '.tex && dvipdfm ' rEname ' && ' rEname '.pdf'];
    system(cmd);
end
end

[Adam Danz]

It's probably a lot easier to just use formattedDisplayText which was not available when my answer was originally added.

Answer 3

[Marina Fernandez]

Another option is to convert the model to text, separate it by rows and convert that to a cell in order to write the results of the model in excel format:

text_model = evalc('disp(model)');

split_text_model = split(text_model,char(10));

xlswrite(excel_path,cellstr(split_text_model),'results','A1');

[Elsa Fouragnan]

this doesn't work on my end.