Sign in to follow this  
Followers 0
marksilvert

stats

9 posts in this topic

getting confused, how do you know if something is parametric data or non? What is parametric data!!

Share this post


Link to post
Share on other sites

Numerical data can be continuous or discrete.

Now, continuous numerical data may or may not follow a parameter. If it follows a parameter, it is parametric and if it does not, it is non-parametric. What is that parameter, I shall try to explain in my next post.

Share this post


Link to post
Share on other sites

its parametric if it follows a normal distribution. this is what we use to describe most things in populations - like height, weight, blood sodium etc.

if you draw a bell shaped curve, we know that most people will be scattered symmetrically around the mean, and that 95% of us will lie plus or minus 2 standard deviations around that mean. so if we want to do some stats (as you do...), we would use a parametric test, which assumes this symmetrical distribution (and is therefore a bit simpler).

but some things are not normally distributed - for example, if the data is skewed one way or the other, or if there are 2 peaks - like schizophrenia incidence in women with increasing age. then you need to use non-parametric tests.

hope that makes sense!

Share this post


Link to post
Share on other sites

Thanks for explanation fibi.

Furthermore if either question asks you to compare or otherwise correlate the means or you know that means can be calculated for this type of data and they don't say that the data is non-parametric or skewed, then it is most likely to be parametric. So comparing (or otherwise correlating) means is the keyword.

Non-parametric test compare or correlate medians or proportions.

If it is median, go for non-parametric tests, and if it is proportions or rates, go for a chi-square or McNemar.

Share this post


Link to post
Share on other sites

im still totally confused can you explain with examples, what the types are, from parametric and non parametric and also categorical, im so screwed for wednesday!!!

Share this post


Link to post
Share on other sites

All I can say is that telephone me and I shall try to explain it to you as best as possible.

Share this post


Link to post
Share on other sites

markymark, which book are you using?

just because some make it much more complicated than it needs to be.

all you need to know is

1. categorical - are things that you can't give a number - like marital status. so the categories will be married, single, divorced, etc.

or like 'dead or alive' - only 2 categories, so you call it binary, but it's still categorical, when you need to work out the stats.

Some things you can have a number for, but the number doesn't really mean a number - it just stands for something - like social class - the nos 1-5 are used but they don't really mean anything numerically so we still consider them categorical (but because they are in an order, we call them ordinal)

Everything else is quantitative data - meaning it's numbers! if your numbers follow a normal distribution (as discussed earlier), you use parametric tests. if not, use non parametric tests.

the only thing you need to do now is learn exactly which tests to use for which groups of data - because it is different if the data is paired (ie. from the same sample before/after an intervention for example), or unpaired (for example from 2 separate groups). do this by making youself a table and just memorising it.

good luck. i'm sure it will just click very soon!!

Share this post


Link to post
Share on other sites

and anyway, even if you can't understand it, there will only probably be maximum 1 emi or bof question on this specific thing. if you guess you will get some right anyway!

Share this post


Link to post
Share on other sites

to be very honest....just try memorizing the table given in B notes...I think that works best.

Share this post


Link to post
Share on other sites
Sign in to follow this  
Followers 0