Variable Transformations in SPSS: Square root (sqrt) and Lg10

A quick tutorial on how to perform a square root or Lg10 transformation on positively skewed data in SPSS

[queencurlz]

Just wanted to say thank you much. I literally just had a panic attack about this huge assignment I have due on Monday and this video really helped remind me what the heck I am suppossed to do. Thank you much!

[SiobhanPhD]

You are so welcome. I hope this made the assignment a bit easier adn perhaps decreased your hatred of stats by a smidgen. :)

[SiobhanPhD]

Hey H Ruso-

Sometimes a variable is so far from normal there simply isn't a way to transform it to bring it close to normal. In cases like that the best idea is to graph your data to get a real clear idea of what you have and then often you'll need to group the data and run a non-parametric stat instead. Not the best option but sometimes your data will dictate that.

[SiobhanPhD]

Swan - Those numbers aren't bad.  Partially it depends on your field and what you are measuring.  In psych I'd be perfectly happy with those numbers but if I was looking at say blood glucose levels... I might be less thrilled.
I frequently go with below .01 on Shapiro-Wilk as it is overly stringent in my opinion if you use a .05

[SiobhanPhD]

One other idea - graph it and make sure it's positively skewed. If it was negatively skewed these transformations will make it worse. If it is simply neg skewed then flip it - I have instructions in either this vid or another in that series -and then apply the transformation.

[ssd]

Once a variable has been transformed, assuming that the skewness is improved, can it be included along with other non-transformed variables in a MANOVA? Thanks.

[msmeenabhatta]

How can you transform negative values (less than 0)?

[ibrahimsaid]

why we use sqrt instead of exp or trunc? what are the rules of choice please? if my data still non-nromal after applying all functions what should I do?

[hruso]

Hi there. I am quite new into stats. Thank you for your demostration, it was very helpful. But my data is still not normally distributed after transformation. Do you have anything else in mind. My data is about a drugs effect on human sperm motility. Other scientists used arcsin before analysis but it apperars that it doesnt work on my data. Can you please help?

[hollyw]

What do I do, if I have tried a Log10, a Squareroot and a reflectionLog10 transformation but they still aren't correcting my data?

[swanpyaeoo]

Hi, thanks very much for the video. But I still have a problem. My data is still non-normal even after transformation. However it become 0.043 according to Shapiro-Wilk test which is close to 0.05. However, according to Kolmogorov-Smirnov test, it is not significant at .200. In this case, can I assume my data is normal or not? Please help me.