I have tried to compile information as to how examine various mouse issues using Paint and/or some mouse count logger as described by outerspace.
Feel free to contribute and discuss.
Here is what I found (for some reason I was too lazy to capitalize, bear with me
prediction/angle snapping: draw straight lines (horizontal/vertical) at high speed. see if they are perfect or real. try to find maximum prediction free speed.
interpolation: draw straight lines (horizontal/vertical) at medium-high speed. see if small steps are done as one pixel.
jitter: draw diagonals at low-mid speed. see if there are steps.
malfunction rate: (might want to lower windows sens) move horizontally at mid-maximum speed. see if cursor follows or jumps. try to find maximum speed w/o abnormal behaviour.
acceleration: (might want to lower windows sens) move mouse an exact distance at an exact and constant angle - once at low, once at high speed. see if cursor moves the same distance. try to find maximum speed w/o acceleration.
tracking precision: specify mouse reference point. move mouse away, draw circles, return to reference point. keep the angle of the mouse as steady as possible (avoid turning the mouse). see if the cursor returns to the same spot. try different speeds.
LOD: put CDs between mouse and surface left and right of the sensor. determine number of stacked CDs needed to stop tracking (1CD ~1.2mm).
z axis tracking: lift and return mouse without moving it. see if cursor stays reasonably in one place.COUNTS PLOT
prediction/angle snapping: unnaturally constant count rates (vertical line: x=0; horizontal line: y=0).
interpolation: count rate > 1 on non-moving axis (vertical line: x-axis; horizontal line: y-axis)
jitter: count rates > 1 at low speeds?
malfunction rate: quick swipes/flicks will show as huge bumps/hills in the count rate. when malfunction rate is hit, the count rate _usually_ drops to zero somewhere on the ascending slope).
acceleration: if you move solely on one axis, you can sum up the counts of both runs and see as to how they coincide or differ.
tracking precision: convert all x and y counts so that they inculde their absolute (turn minus to plus). then calculate the distance for each sample using: d = sqrt( x^2 + y^2 ). if you sum up all distances, you will get the travelled distance. now if you sum up all the original x counts and y counts seperately (without taking the absolute) and then calculate d_error = sqrt(sum_x^2 + sum_y^2), you should get a smaller value (ideally zero). 2nd_value/1st_value*100 should give you the difference in percent.
smoothness: move mouse at constant speed (draw circles!). convert all x and y counts so that they inculde their absolute (turn minus to plus). then calculate the distance for each sample using: d = sqrt( x^2 + y^2 ). the d rate (distance travelled per time aka speed) should be constant.
edit: Let me know if this should rather be a separate thread.