0

I am having quite a few problems with transforming a set of data with values between -1 and 0, as I need to normalise them.

I tried to use the following formulae:

[(−min())/(max()−min())]

AND

atahn[(−min())/(max()−min())]

But both don't seem to work... As the p-value is always below 0.05 using the Shapiro-Wilk test.

Can I transform the data using the log[(−min())/(max()−min())] formula? Is this appropriate? Thanks xx

CLARIFICATION: Sorry guys, I am quite new to these things, I thought "normalise" meant making sure that a set of data was normally distributed. I have a set of data and I have to check to see whether they are normally distributed - and to see that, I was told that I have to use the Shapiro-Wilk test, which must be above 0.05, a histogram, and a Q-Q graph (I am using SPSS). But the data that I have comprises all negative values, and when I checked they were not normally distributed. On the Internet, I read that I have to log-transform my data, but since they're all negative values, I tried the two options above in my original post - plus log(x+1) yesterday evening. But nothing seems to work. My question was whether I could use the following formula log[(−min())/(max()−min()) to transform my data and achieve normal distribution? Because I can't find anything else on the Internet on that... Or if there was anything else I could try?

sunnysonny
  • 1
  • Please explain how normalization produces a p-value. It seems like you might be trying to do something very different from what you are asking about. – whuber Apr 08 '22 at 18:51
  • 1
    “Normalization” does not produce normal distributions. Normalization produces values in the interval $[0,1]$. For what reason do you need to “normalize” your data? – Dave Apr 08 '22 at 18:54
  • Sorry, I didn't know that "normalize" and achieve "normal distribution" were two different things. I edited the post - hopefully it's makes sense now. – sunnysonny Apr 09 '22 at 18:11
  • What do you mean when you ask if you can do it? You did the transformation, right? So you can do it. If you’re asking if you should do it, assess the results. Do the transformed data look normal? It might help to say more about what you’re doing, particularly why you want a normal distribution at all. – Dave Apr 09 '22 at 18:47
  • I need a normal distribution because after that I am required to do an ANOVA. I just tried using the formula log[(−min())/(max()−min()) and it does look like a normal distribution now. But I'm not sure whether that's an appropriate way or if I'm just making up the results, because I couldn't find anything on the Internet regarding this method... – sunnysonny Apr 09 '22 at 21:37
  • Please ask about the actual statistical problem you have. Your current question is not going to solve it--it will instead lead you astray. ANOVA does not require normality. When you apply a nonlinear transformation you are changing what ANOVA does, anyway: it no longer compares arithmetic means of groups. If you really, really, want to transform anyway, then simply negate all your values and watch for answers to the same question about values between $0$ and $1.$ – whuber Apr 10 '22 at 15:12
  • This site search turns up many relevant posts. – whuber Apr 10 '22 at 15:43

0 Answers0