How do you confirm the skew of something, is it something to do with the proportion of mean to median?
Thanks
You can, if you're just given these pieces of information.
A positively skewed graph will have the median on the
left of the mean.
A symmetric one (similar to a bell curve/ normal distribution) will have the median and the mean at the same point/ value.
A negatively skewed graph will have the median on the
right of the mean.
Alternatively, on a box plot, you will see this as:
positively skewed - the right 'whisker' will be longer than the left. (or the line from Q3 to the maximum)
symmetric - Q2 will be in the centre, both whiskers are equally long.
negatively skewed - the left 'whisker' will be long than the left.
Of course, you can work it out if you're given raw data as well.
(Source: some of this has been taken from my Further teacher's notes from last year.)
When asked to list what types of transformations would be suitable for a table of data values, should I only base my answer on the 'circle of transformations' suggestion, or compare r-values for more accuracy? One of the transformations in the 'circle of transformations' gives a pretty low r value compared to the other transformations.
I'm not really sure what you meant here (could you provide an example?), but how I interpreted this question, you're asking about questions which ask you what's the best transformation to apply for that set of data.
Well... the first thing is to see the overall look of the data. Is it curved? Which way is it curved?
How would an x
2 or a 1/y transformation affect the data?
Would it have the highest possible r
2 value? (Since the highest r
2 value will provide you with the closest fitting transformation and 'neatest' line - i.e. "more accuracy" in your terms.) (I know I'm explaining this horribly, I'm sorry. I hope I explained it well enough. It's sort of difficult to word it.)