Which stats test to use?

 Which stats test to use?
Posted: Saturday, March 18, 2017 @ 10:29 AM
avatar
Joined: 11 October 2009
Posts: 187

Help ……I’m having a crisis…….within our case study on learning in order to do the required stats test I am asking students to choose an emoji happy face or sad face to demonstrate their approval/disapproval of visual/verbal/organisational mnemonic strategies. I thought that Chi Sq would be appropriate as happy/sad is nominal data and we are looking to see if there is a difference - but I can’t decide if this would be Independent groups design as all students are doing the same thing?


Posted: Saturday, March 18, 2017 @ 11:35 AM
avatar
Joined: 8 December 2008
Posts: 714

What do you mean by all students are doing the same thing?

Posted: Saturday, March 18, 2017 @ 12:12 PM
avatar
Joined: 11 October 2009
Posts: 187

Sorry I meant that data would be related i.e. participants could have a score in more than one cell i.e. happy with visual, unhappy with verbal and happy with organisational. I have changed the hypothesis to - there will be a difference in males and females preferred mnemonic i.e. use of visual techniques or use of verbal techniques. So this will satisfy the requirement for Chi Sq of data being nominal (verbal or visual) it being a test of difference (males and females) and requiring unrelated data (males and females as separate groups)?


Posted: Saturday, March 18, 2017 @ 04:51 PM
avatar
Joined:
Posts: 97

Hi, wouldn't it be independent groups as you are using males and females?

Posted: Saturday, March 18, 2017 @ 07:04 PM
avatar
Joined: 1 November 2007
Posts: 1987

I think chi-square as you have a 3x2 box (3 choices of mnemonic and 2 genders)

Posted: Saturday, March 18, 2017 @ 09:18 PM
avatar
Joined: 2 May 2008
Posts: 915

No - each person is represented in more than one box because they each vote for each type of mnemonic - don't they. With chi square you have to imagine that each person is only allowed to stand in one box so it would only work if you are comparing whether males and females prefer a specific type of mnemonic.

Posted: Sunday, March 19, 2017 @ 04:02 PM
avatar
Joined: 1 November 2007
Posts: 1987

Cara... Surely if the IV is make / female it is independent groups.? What other test would you use?

Edexcel set this on their SAMS and then asked students to complete a chi-square test.

"A group of students have been having difficulties in managing their anger and
were advised to attend an anger-management programme. They completed an
anger assessment before and after attending the anger-management programme.
Researchers wanted to see if there were gender differences in the effectiveness of
the programme. The anger assessment results for males and females are presented in table 2"

This is independent measures (male and female) BUT the participants still had a score in each box as they completed a questionnaire before and after treatment. Should this not be a chi-square then?

Posted: Sunday, March 19, 2017 @ 06:19 PM
avatar
Joined: 11 October 2009
Posts: 187

If the anger levels were scored wouldn't this then count as at least ordinal data - so a Mann Whitney U would be appropraite?

Posted: Sunday, March 19, 2017 @ 09:10 PM
avatar
Joined: 1 November 2007
Posts: 1987

No as they were looking at gender differences not levels of anger before and after treatment. The question was asking them to complete a partially filled table working out chi-square

Posted: Sunday, March 19, 2017 @ 09:58 PM
avatar
Joined: 2 May 2008
Posts: 915

For chi-square the data in each square has to be independent. I'd have to look at what they have done - where is the question on the Edexcel SAMs?

You could have males who scored more or less than a certain figure (i.e. they benefitted or didn't) and females who scored more or less.

Posted: Monday, March 20, 2017 @ 10:15 AM
avatar
Joined: 1 November 2007
Posts: 1987

It was on the first set of A level SAMS, pg 74 (Paper 2, Section B Q7)

I am really confused about chi-square now :( I have used this for an observations looking at helping and non helping behaviour but we simply recorded the number of helping behaviours and non helping behaviours shown by males and females. However as we did this in the school cafe the same students could have shown both behaviours. Does this mean I shouldn't have carried out a chi square?

Posted: Monday, March 20, 2017 @ 10:20 AM
avatar
Joined: 8 July 2009
Posts: 149

I'm thinking along the same lines as you are Jenny...we do an observation on the school field focusing on gender differences in the type of 'play' (active/passive). We do a tally and of course one person could display both types of 'play' but that doesn't matter because we're looking at gender differences so it is chi squared. I hope!

Posted: Monday, March 20, 2017 @ 10:48 AM
avatar
Joined: 2 May 2008
Posts: 915

It does matter. In a 2 x 2 contingency table there are 4 squares. Each person can only be represented in one of the 4 squares so you could not score, for example, one female in two squares. That's what the 'independent' means in this context.

However if the rows were male/female and the columns were high score/low score then each person would only be tallied in one of the four squares.

Posted: Monday, March 20, 2017 @ 11:35 AM
avatar
Joined: 1 November 2007
Posts: 1987

Ok thanks Cara.

Posted: Monday, March 20, 2017 @ 11:53 AM
avatar
Joined: 2 May 2008
Posts: 915

Chi-square is frequency data - the data in the squares is a frequency count of how many people were in that group, and a person can only be in one of the four groups.

If I wanted to know whether males or females were more helpful, I would get a score for each person of how many times they helped (or a percentage of how many times they helped out of a total number of observations) and then do a difference test for independent groups. Your two groups are categorical data (males and females) but it is the scores/data for each group which is key in deciding what test. You are actually looking at the difference in scores i.e. a test of difference with possibly interval data.

Or you can reduce the scores to nominal data (high or low and male or female).

Posted: Monday, March 20, 2017 @ 12:33 PM
avatar
Joined: 8 July 2009
Posts: 149

Yep - my rows and columns are as you've said Cara so I'm doing it right!

Posted: Monday, March 20, 2017 @ 02:08 PM
avatar
Joined: 1 November 2007
Posts: 1987

Thanks Cara