 |
|
|
|
 |
 |
|
|
 |
|
||
 |
 |
|
|
 |
|
||
 |
|
||
 |
|
||
 |
|
||
 |
|
||
 |
|
||
 |
|
||
 |
|
||
 |
|
||
 |
|
||
 |
|
||
 |
 |
|
Six Common Types of Bad Plots
(With all due respect to Matt Groening)
 |
 |
| The Handmade Horror | The Bar Monster |
|---|---|
| Pros: Quick and easy, gives you a nice preview of the data's behavior. | Pros: Snazzy look is sure to impress fellow students. |
| Cons: Usually too small for accurate plotting of the data, and uneven lines make interpolation between points difficult. It may also be covered with illegible notes. | Cons: Snazzy look will not impress your teacher; bar graphs make it more difficult to display error bars and fits to your data. This one has the even more basic problem that the voltage data are plotted against row number instead of the radius. |
| Warning Signs: Smudgy pencil lines, slanty error bars, unlabeled axes. | Warning Signs: Data points lie on integer x-coordinates, relationship between variables looks very different from what you expected. |
 |
 |
| Error Flynn | The Stealth Bar Monster |
| Pros: Flamboyant, daring error bars make it look like you've been suitably cautious. | Pros: Elegant and scientific-looking. |
| Cons: Did you at any point tell the computer what the uncertainties on your data were? If not, these error bars almost certainly don't represent anything relevant to your experiment. These error bars are 10% of the y-coordinate, a common default setting. | Cons: This plot contains a subtle but fatal error, usually made by students using spreadsheet programs to plot their data. The x-data (radius) has been used as a label for the y-data rather than as an independent variable. If you look carefully you can see that the x-axis isn't scaled uniformly, which will certainly distort the apparent function V(r) badly. |
| Warning Signs: Suspiciously uniform error bars, or error bars which seem much larger or smaller than the scatter in your data. | Warning Signs: This problem can be very difficult to spot, especially when the intervals between points on the x-axis are similar but not equal to each other. For example, the x-data shown above are spaced at intervals of 1.1, 1.2, 0.9, 1.0, 1.1, 1.1, 1.0, and 0.9 cm. On the plot each point is spaced by the same distance, which distorts the appearance of the data only slightly. Be careful to select the "scatter plot" option when using programs to graph for you; this will plot the points on a scaled axis in both directions. |
 |
 |
| Jaggedy Ann | Out Of Line |
| Pros: Straight lines connecting your data points make it easier to interpolate values. | Pros: Best-fit line to data is an attempt to find a useful model for the behavior of your data |
| Cons: Lines can obscure the data points and any attempts to fit a curve to the data. If the points are very close together, then the lines can be easily mistaken for an attempt to fit your data with a function. | Cons: Although a linear relationship between two variables is common, it is by no means universal. You shouldn't plot a linear fit unless it is justified. If the points look like they are following some other relationship, try fitting different kinds of curves to it (e.g. exponential, quadratic, or sinusoidal) and plot the one which seems to work best. You should never display a bad fit to the data on your plot. |
| Warning Signs: Avoid options labeled "interpolate" or "spline" unless you make it clear that the curve drawn on your plot is not a fit. | Warning Signs: If your data points do not seem to be randomly scattered about the line, or the line misses more than a third of your points' error bars, then try another kind of curve. |
Created by Ben Mathiesen
Last Updated by Andy Pawl
Sept. 2003
apawl@umich.edu