Anyone know their stats?? PLLEEEASE help??

[treedelivery]

DH keeps asking me for help and I can't get my head around it.

Ok - dh has to study the non verbal communications of men versus women within a couple. He has watched 5 couples and counted how many times the man touched the woman and visa versa. He also noted where they touched.

Ok - so 5 couples you chose because they were known to you and coming to your house - would you called them paired samples because you are observing between couples, or are they inpendent samples??

I think the date he collected in ordinal, so is it non-parametic or parametric? We can't get to a conclusion and the more we talk abut it the more confused we get.

Please help!?!?!?!?

I assume they knew what he was doing!

Yes of course. He was countng how many times they touched hands or shoulders etc.

Wolfram Alpha is my maths bible might be able to help. Have you looked at it?

Thank you empirestateofmind. I think he'd need to have a good look at it as an initial search by me generated www.wolframalpha.com/input/?i=parametric+data which I found rather scary

I am a statistician so will give you my quick 5 mins!

You have written about several different concepts but I'm not sure what you are aiming to achieve overall, what is your hypothesis?

Yes this would be paired data as you have variables relating to the men and the women but the men and women are couples so the variables are not independent. But the sample of men and women chosen is not independent as the couples were not chosen at random from an address book for example. You are confusing independence of variables and observational units.

If you have ordinal/ranked data, then in this case you might use a non-parametric test but how are the data ordinal?

Parametric data implies you are assuming a particular distribution with certain parameters (e.g. normal distribution), whereas non-parametric data assumes no particular distribution.

It would help to know what exactly you are trying to study here. Hope some of that helps.

I agree you need a hypothesis first.

I would have argued that this was an opportunity sample more than any other kind ( any pair who was willing and able to participate in the investigation.

It will certainly be a non parametric test. Possibly chi squared ( which would use nominal data - that is data divided into categories ( touched shoulders, neck, arms etc). If the hypothesis and data well defined enough you might get away with a spearman rho (correlation) or even a Wilcoxon.

There is too little information in what you offer.

Re reading. Is this psychology?

If so I think you will find its
a) opportunity sample
b) repeated measures
c) ordinal data
d) Wilcoxon test should be applied.

Still the hypothesis would help confirm that.

gosh my head is spinning. good luck with it all

Oh bless you all, if only I'd been able to stick about to grab you! The dd's came home from a week away so just had to go but thank you for your input and I hope you check in again.

He has a hypothesis - there will be a significant difference in the amount men and women touch.
The null hypothesis being there will not be a significant difference between men and women.

Dh reckons that is two tailed?

He has observed each couple for an hour and tallied up how many touches and where. Only 5 couples as this isn't a huge massive piece of research just a student project. Plus we haven't that many friends

So he now has a table of how many times the man touched his wife on the hand and visa versa, on the shoulder and visa versa - and so on. Plus some lovely bar graphs etc.

inthecloud - I don't really understand what the term 'independent' refers to. Is it the subjects or the variables? So I'm not a whole load of use to him and I feel really sorry for him as he's been at it for days hours.

Hi,

Yes, it is two tailed. Predicts a difference without direction.

If you use the hypothesis ( experiemental or H1 or alternative, whichever word you use) men tough significantly more than women ( or vice versa) that would be a one tail) Null is always there will be no difference ( or significant difference as stated here) .

I think on that hypothesis you might well be best advised to go for a chi squared anaysis ( or series of them varying the hypothesis each time) .

A straightforward male female x amount of touch cells should do it.

Your sample is very small so a contingency table would be best.

You cannot use parametric testing because of the sample size and liklihood of a skewed data distribution. Besides your data is nominal ( category) as set up by your hypthesis.

Independent measures means you have participants in your sample who have been dvivided such that they work on two different tasks ( or more).

Repeated means you use the same group of Participants but engage them on different tasks serially.

In your case, this does not need to apply if you use a contingency table ( Chi squared).

You have an opportunity sample ( anyone available and willing to participate) .

Your data is nominal because it is divided into categories ( male and female on the one side - looking for a difference in amount and pattern of touching) and the amount and pattern of touching on the other ie cells of shoulders, arms , hands etc.) or you could amalgamate the data, do a four cell analysis and just have male and female and amount of touch ( add up all touch to each part of the body) and do a 2x2 chi ( dont forget yates correction if you do that). Or do both. I would do both.

Variables: the things which you manipulate and measure ( IV and DV respectively). Your DV here is amount of touch.

Your significance level should be 0.05 ( p > 95% confidence) because in psychology thats convention but in maths that would give you the best chance of not obtaining a false positive result - which means you have only five chances in 100 of saying there is a difference where there is none. It balances type on and type two errors.

Hope that helps. if he has a chart of touch and male female already , the chi is ready to go just put in the totals and apply the formula. If you do not have the formula I will need to get my stats book out as I can never remember it but you can usually find it anywhere including the internet or most books ( especially psychology books)

Hello again,
It would be two tailed. One tailed would be if you were expecting men to touch more than women, or vice versa (i.e. have a greater or less than hypothesis).

I agree you could do a contingency table/chi square analysis.

Independence can refer both to the subjects and the variables.

The other posters here seem to be specialists in psychology or something similar so would go with their advice.

I dont want to mislead anyone. I am a statistician but I have taught statistics to psychologists on occassions.

I could actually cry with relief. I dreamt about this last night

jabed and intheclouds00 - thank you. Seriously.

So what happens? He does the chi square and gets a number and...so what? How does that prove his hypothesis or not?
Don't get me wrong he has course materials and we have both tried to read around the subject but if you can't grasp the basic message all the extra reading makes it harder. Having someone talk it through in terms of your own project makes in so much easier.

He has the discussion planned out already, he is using the four cell analysis you talk about jabed [how much men and women touch] as the main finding to discuss but that to look at the pattern of touch would be intersting and warrants further investigation. That great disclaimer. His main finding is that people are highly unlikely to touch eeach other on the leg at a bbq as they aren't sat down to do it

dreamt???

wow another statistician on mumsnet!! :-)

maybe we know each other jabed..

OK, so you put in the figures on the contingency table and out comes a figure.

That is your observed ( or calculated) chi value.

You then calculate the degrees of freedom. If you are using a 4 x 4 contingency this will be 1. It always is.

You have set your P > 95% ( 0.05 significance level or written usually as
SL= 0.05)

Now you need to get a set of statistical tables. In there will be a table for the values of Chi squared.
You may find one online. Most A level psuch books also have them as will most statistical books. You look up the relevent values for 0.05 ( two tail) and 1 dof ( degree of freedom).

That will give you the table value ( or critical value of chi squared)

You compare your observed value to the critical value. If the observed value is greater than the critical value then your result is significant and you reject HO ( null hypothesis) and accept H1 and make your decision accordingly.

Now, maybe I should not do this but , I have a table of critical values of chi to hand and the figure from the table ( ie critical value of Chi for 1 dof and in a two tailed test is 3.84 .

If the calculation of Chi done by your dh is greater or equal to that nuber then his result is significant and he can say that there was a significant difference in the amount of touching done by men and women.

If it is less than 3.84 then he has to reject the H1 and say that there is no difference.
Hope that helps.

In the clouds)) , its a small world, maybe we do. I dont know.

Two things - if you want to discuss touching you need to complete analysis of chi for each earea observed to establish significance .

If your results are not significant you can only discuss your findings as a series of unsupported ( statistically) observations. Whilst that is not inappropriate in terms of observation , within the context of research methods and statistical analysis it is notevidentially supported sufficently.

You might with some massaging of the stats be able to find a level at which the results , whatever they are do become significant ( do that by looking up your observed value across the tables and then looking at the level of P that is significant for).

I am surprsied anyone can map out the discussion without knowing the results? I take it this was done from the descriptive statistics? So the discussion might have to run along the lines of " whilst not statistically significant ( if its not) there was eveidence of more / less touching in areas such as ....."

You could do a chi on the hypothesis that significantly less touching takes place in certain areas and run a 2 x2 contingency on that in additin to that of men and women touching for significantly different amounts.

Critical value for a one tailed test would be 2.71 at p> 95%

You cannot really have a main finding which is not analysed in some way from inferential statistics if you intend it to be accepted as valid and reliable.

jabed I am your biggest fan. Message me anytime you need some one onside on AIBU.

Ok. He did the numbers when he got home from work and is just emailing his tutor now. He got 3.853 so he gets to accept the hypothesis. Cool as he tells me the research he has found already actually says there is no difference in this age group.
But Our bbq test was blatently scuppured by a in lurrve newlywed couple with a very hands on chap. So he has lots of individual points to pull out and chat about - he already did that from the raw data, it was obvious looking at the data on the night.
So ideally now he should test each touch scenario? To look for individual interesting areas.

You know the chi result, is anything over 3.84 significant? Is 3.85 just as significant as 3.86 and so on and so on. Or is this result just significant?

It's really interesting, all of this. My hat goes off to you all for undserstanding the basic maths at the root of it all. It's very impressive.

Thank you, he really needed this help and it's been nice for me to have something positive I can do to help him.

If it is 3.84 and the chi calculated is the same or greater it is significant. Sats doesnt talk in terms of " just" or " very". Besides, with a 95% confidence level , it isnt " just anything. The accuracy is that with gives only five chances in one hundred of the result being incorrectly drawn. Hope that helps.

Thank you jabed. I just wondered if a greater number meant greater significance, some of the chi calculations for touches in different areas are vastly different.

Well, it's nearly finished. What would he have done without you, mumsnet and me as the go between.

Woo Hoo!

BTW, I still think it's very very cool that you understand this maths. This has almost inspired me. I want to know why and it's pissing me off that I don't. Maybe I should go back and study and understand the basic maths at the root of all this. Basic as in fundamental, not as in easy. It sure isn't basic or easy to me anyway!