How can you clean negative values in a logarithmic scale bar chart?

1 Clipping method

The clipping method involves setting a lower bound for the negative values, such as -1, and replacing any value below that with the bound. This way, you can avoid the error of taking the logarithm of a negative number, but you will also lose some information about the magnitude of the negative changes. To implement this method, you can use a conditional statement in your data processing tool, such as pandas in Python. For example, if your data frame is called df and your column of interest is called value, you can use the following code to clip the negative values: df['value'] = df['value'].apply(lambda x: -1 if x < -1 else x)

2 Transforming method

The transforming method involves applying a function to the negative values that converts them to positive values while preserving some information about their relative size. One such function is the inverse hyperbolic sine, which is defined as: asinh(x) = ln(x + sqrt(x^2 + 1)) This function has the property that asinh(x) is approximately equal to ln(-x) when x is large and negative, but it also works for small and positive values. To implement this method, you can use a built-in function in your data processing tool, such as numpy in Python. For example, if your data frame is called df and your column of interest is called value, you can use the following code to transform the negative values:

import numpy as np
df['value'] = df['value'].apply(lambda x: np.sign(x) * np.log(np.abs(x) + np.sqrt(x**2 + 1)))        

3 Comparing methods

After applying either of the methods, you can create a logarithmic scale bar chart using your preferred data visualization tool, such as matplotlib in Python. You should compare the results and see how they affect the visual representation of your data. For example, using the clipping method, you will see a uniform bar height for all the negative values below the lower bound, which may obscure the variation among them. Using the transforming method, you will see a more gradual decrease in bar height for the negative values, which may reflect their relative difference better. However, you should also be aware that the transforming method may distort the scale and make it harder to interpret the actual values.

4 Communicating methods

Whichever method you choose, you should communicate it clearly to your audience and explain why you used it and what are its implications. You should also provide a reference scale or a legend that shows how the original values are mapped to the logarithmic scale. You should avoid misleading or confusing your audience by using labels or annotations that are consistent with the method. For example, if you use the clipping method, you should not label the lower bound as zero, but as the actual value you used. If you use the transforming method, you should not label the bars as percentages, but as the inverse hyperbolic sine values.

5 Here’s what else to consider

This is a space to share examples, stories, or insights that don’t fit into any of the previous sections. What else would you like to add?