plotAverageFreq(varargin)

function outText = plotAverageFreq(varargin)
persistent filterLength
if ~nargin
    outText = 'Plot Average Frequency';
    return
end

    analysisAxis('eventFreqPlot', varargin{4});
    
    handles = get(gcf, 'userData');
    whichAxis = find(handles.axes == gca);
    
    % get handles to the event traces
    events = getappdata(gca, 'events');
    
    % plot the frequency data
	xData = handles.minX:handles.xStep:handles.maxX;    
    if isempty(filterLength)
        % suggest a window that has, on average, 5 events in it
    	filterLength = round(length(xData) / length(events(1).data) * handles.xStep * 2.5) * 2;
        howManyDigits = floor(log10(filterLength));
        filterLength = round(filterLength / 10^howManyDigits) * 10^howManyDigits;
    end
    tempData = inputdlg({'Boxcar length (msec)'},'Plot Freq...',1, {num2str(filterLength)});      
    if numel(tempData) == 0
        return
    end
    filterLength = round(str2double(tempData) / handles.xStep);
    if filterLength > size(xData, 2)
        filterLength = size(xData, 2);
    end
    yData = zeros(numel(events), size(xData, 2));    
    for i = 1:numel(events)
        if numel(events(i).data) > 1
            % this is simply a boxcar filter, but implemented using the
            % filter command it took 100x longer
            changeData = ones(1, 2 * numel(events(i).data));
            changeData(end/2 + 1:end) = -1;
            whereData = [round((events(i).data) / handles.xStep - filterLength / 2) round((events(i).data) / handles.xStep) + filterLength / 2 + 1];
            [whereData indices] = sort(whereData);           
            lastSum = sum(whereData <= 0);
            yData(i, 1:whereData(lastSum + 1)) = lastSum / (filterLength * handles.xStep / 1000);
            for j = lastSum + 1:min([find(whereData < length(yData), 1, 'last') - 1 size(whereData, 2) - 1]);
                lastSum = lastSum + changeData(indices(j));
                yData(i, whereData(j):whereData(j + 1)) = lastSum / (filterLength * handles.xStep / 1000);
            end
        end
    end
%     patch([xData xData(end:-1:1)]', [mean(yData, 1) + std(yData, 1) / sqrt(numel(events)) mean(yData(:, end:-1:1), 1) - std(yData(:, end:-1:1), 1) / sqrt(numel(events))]', [.8 .8 .8], 'edgecolor', 'none', 'parent', handles.analysisAxis{whichAxis}.eventFreqPlot);
    line(xData, mean(yData, 1), 'color', [0 0 0], 'linewidth', 1, 'parent', handles.analysisAxis{whichAxis}.eventFreqPlot);
    line(xData, mean(yData, 1) + std(yData, 1) / sqrt(numel(events)), 'color', [1 0 0], 'lineStyle', ':', 'parent', handles.analysisAxis{whichAxis}.eventFreqPlot);
    line(xData, mean(yData, 1) - std(yData, 1) / sqrt(numel(events)), 'color', [1 0 0], 'lineStyle', ':', 'parent', handles.analysisAxis{whichAxis}.eventFreqPlot);        
    set(get(gca, 'userData'), 'string', ['Boxcar filtered at ' num2str(filterLength * handles.xStep) ' ms']);            
    ylabel(handles.analysisAxis{whichAxis}.eventFreqPlot, 'Hz');
    
    set(handles.analysisAxis{whichAxis}.eventFreqPlot, 'ylim', [0 3 * max(get(handles.analysisAxis{whichAxis}.eventFreqPlot, 'ylim'))]);
    filterLength = filterLength * handles.xStep;