Plotting averages when there are missing values I have count data for a number of subjects in different groups I would like to compare. The averages of the cumulative sums are shown in the figure below. As you can see the Red group has a spike at the end. This is because one subject in that group had fewer trials. This subject had the lowest cumulative number of counts for the previous trials and was pulling down the average.

How would you plot and interpret data like this? I wish to avoid imputation or other techniques "fancier" than calculating simple statistics, ideally I would plot only the data itself. I already have a plot containing each individual, but also want a group comparison.
Edit:
The y axis is cumulative # of successes while learning a new task at each block of trials. The x axis should probably read "Block #" rather than trial number.
The successes were followed by a reward, so cumulative number of successes is also a measure of total reward received over the course of the study, which may be the outcome the subjects were attempting to attain. 
Alternatively the learning can be measured as the number of successes per block over time. Some subjects may learn slower but end up with higher final performance and thus appear "better", however subjects that used a "quick and dirty" strategy may be maximizing the total reward attained despite their apparent lower skill level.
 A: Sometimes we try too hard to put the entire message in one graph. You might benefit from pairing the cumulative view with a non-cumulative version. Here's a rough mock-up with similar data.

The (lower) data view can make it easier to see certain patterns. For example, cumulative lines with the same slope will be represented as data values at the same level. And since each X is really a block of values in your case, you might consider adding some indication of that in the data views, such as with error bars or box plots.
A: It sounds like you aren't looking to do any statistical test, you just want a way of illustrating the data. If I were you, I would just show the plot as is. You can still see the trends very nicely and the errorbars will likely dwarf the "red spike". You can mention the fact that one subject in the red group had fewer trials in the figure legend.
If you want to do something more fancy, I think data imputation is the only way to go. Essentially, extrapolate your curve fit for that subject to trial #40 and put that into this dataset. I'm not sure that would really change the qualitative result, though, begging the question as to why bother in the first place.
Even if you are doing a statistical test, I'm not sure anything you do about the last data point will matter much.
A: Why don't you simply hide the last value on the red curve? You can comment in the legend that one of the subjects in the red group had only 39 trials instead of 40, so red average is shown only until 39. Given your curves (that are very smooth and almost linear in the end), I think it is perfectly fine to do that.
A: I am not really sure I understand why having less trials by one subject would cause an upward spike near the end. I cannot imagine the circumstances which would produce this graph.  Perhaps in the future you could provide some data (if only mock data) which could demonstrate the situation you face.
From what I understand, I would recommend either that you no be concerned with that last spike on the graph since it is not particularly large and likely could easily be explained through standard unobserved variation.  The overall take away with the blue line is much higher than the other two but otherwise they seem parallel.  If you have sufficiently sparse information such that one individual could so extremely disrupt your graph then I don't think you have anything to worry about since such a slight tweak near one tail is unimpressive.
