2

I am doing some explorative work on two large datasets. One from 2001 and one from 2018. The dataset consists of measured soil-parameters and it contains lots of zero's.

From the transformations done in 2002 on the 2001 dataset, I have some notes written on the sideline of a LOGtransform graph: 'dataset contains '0'-values, 2 is highest number in column, so LOG(data+0.1)'.

The result of this transformation: enter image description here

My Question: Where does this 0.1 come from in relation to the 2 (highest number in column, dataset)? What is this transformation called and what is the idea behind this kind of log transform?

I have the same skewness in my 2018 dataset, but the LOG(data+0,1) does't work as well. Also, can I compare the two with Repeated Measures ANOVA when the two log transformations have different numbers added to the data?

Stevestingray
  • 213
  • 3
  • 8
  • Why do you want to take the logs? What exactly is the problem? – Tim Apr 04 '19 at 14:13
  • I want the data to have a (close to) normal distribution for parametric variance testing of the means. i.e. did soil parameters change significantly over years. – Stevestingray Apr 04 '19 at 14:15
  • Also the following are worth checking https://stats.stackexchange.com/questions/30728/how-small-a-quantity-should-be-added-to-x-to-avoid-taking-the-log-of-zero and https://stats.stackexchange.com/questions/18844/when-and-why-should-you-take-the-log-of-a-distribution-of-numbers and https://stats.stackexchange.com/questions/298/in-linear-regression-when-is-it-appropriate-to-use-the-log-of-an-independent-va/3530#3530 – Tim Apr 04 '19 at 14:18
  • This is answered in here: https://stats.stackexchange.com/questions/30728/how-small-a-quantity-should-be-added-to-x-to-avoid-taking-the-log-of-zero – Tim Apr 04 '19 at 14:34
  • 1
    As far as the name goes, John Tukey called this the "started log." However, I don't think that has gained much popularity, so I would be reluctant to use this term without first defining it. – whuber Apr 04 '19 at 20:55

0 Answers0