If your data are x,y and yh the predicted values, then
scatter(x,y) % the data circles
hold on
plot(x,yh,':') % the regression line
arrayfun(@(x,y,yh) line([x x],[y yh],'linestyle','-'),x,y,yh) % the stem lines
NB: the x,y,yh in the @(x,y,yh) portion of the anonymous function are dummy arguments substituted into the function definition by reference; the actual data are the three arrays as given as the arguments to arrayfun.
doc arrayfun % for more details and anonymous functions.
Takes same number of plotting commands as S-Strider's solution but dispenses with the duplicated data for the lines data array at the expense of a little complexity in arrayfun
