Natural Log

#1
Hi!

I am doing Biomedical studies in proteomics. As I am comparing groups by spot intensity using a computer software ,which is pretty sensitive, and gives me a large no. (>1000) I was told to log it to get a normal distribution, which I did a natural log by excel.

However, a lot of these data is 0 which gives an infinity when u log it.

At the same time, I realise if I use 0 to replace those infinity, these 0 points are usually outliers and I have no idea whether to remove then as a result reducing my n. By doing so, I actually get a statistical significance using 1 way ANOVA. My supervisors told me not decrease my n as this is data selection.

But I really wanna some significance :( especially after working on this for so long.
 

JohnM

TS Contributor
#2
yuppy,

Sorry to break the bad news, but you can't "make" significance happen - it's either present in the data or it isn't.

Since your data is highly non-normal, another option you have, instead of doing a transformation, is to use ranks and run the nonparametric version of the 1-way ANOVA, called the Kruskal-Wallis test.

Go to this link to learn the method:

Kruskal-Wallis 1-way ANOVA