YB
BannerBanner

Stats help (JK2 possibly!)

Author
tristram
Pro-Member
  • Total Posts : 2746
  • Reward points: 1414
  • Joined: 2007/08/01 13:26:18
  • Status: offline
2008/05/12 17:44:58 (permalink)
USA DEALS

Stats help (JK2 possibly!)

i am currently writing my research proposal for next year and ive come up with a null hypothesis. to try and prove it i want to ask two different groups of people three different questions. these two groups are employees of companies and general consumers. the questions are the same for the two groups.

i believe the best way to prove a null hypothesis is using Chi square testing. what i would like to know is a) is it possible with what i want to do, b) can someone explain how you go about getting an expected number. as i understand it, you need observed numbers and expected numbers.
also if there is something ive missed out here in relation to chi squared, please feel free to fill me in

thanks.
#1

17 Replies Related Threads

    USA DEALS
    JK2
    Pro-Member
    • Total Posts : 14595
    • Reward points: 4963
    • Joined: 2003/07/14 15:07:10
    • Location: London United Kingdom
    • Status: offline
    RE: Stats help (JK2 possibly!) 2008/05/12 17:47:26 (permalink)
    Depends on how you code the reponses.

    #2
    tristram
    Pro-Member
    • Total Posts : 2746
    • Reward points: 1414
    • Joined: 2007/08/01 13:26:18
    • Status: offline
    RE: Stats help (JK2 possibly!) 2008/05/12 17:49:57 (permalink)

    ORIGINAL: JohnKerr2

    Depends on how you code the reponses.


    if you mean are the answers yes/no or on a scale etc.

    each question can only provide a yes or no response.
    #3
    JK2
    Pro-Member
    • Total Posts : 14595
    • Reward points: 4963
    • Joined: 2003/07/14 15:07:10
    • Location: London United Kingdom
    • Status: offline
    RE: Stats help (JK2 possibly!) 2008/05/12 17:52:18 (permalink)
    Ok, well the expected number for each cell would be the number you'd expect if there were no differences between the groups. (Usually that means that the expected values would be equal).

    Edit: That's not very clear is it?

    The expected value for each cell in your table is equal to (the total of the row x the total of the column) divided by the total number of observations in the table.
    post edited by JohnKerr2 - 2008/05/12 17:56:46

    #4
    tristram
    Pro-Member
    • Total Posts : 2746
    • Reward points: 1414
    • Joined: 2007/08/01 13:26:18
    • Status: offline
    RE: Stats help (JK2 possibly!) 2008/05/12 17:56:18 (permalink)
    ok i think i understand. what i shall do is knock a table up and then if possible can you advise if im on the right track?
    #5
    tristram
    Pro-Member
    • Total Posts : 2746
    • Reward points: 1414
    • Joined: 2007/08/01 13:26:18
    • Status: offline
    RE: Stats help (JK2 possibly!) 2008/05/12 18:10:13 (permalink)
    What ive managed to do to make it easier is that the correct answer to each question is no.
    so the idea is to compare the number of right answers between the consumers and the companies, to show there is no relation between the companies and the consumers knowledge.

    am i right in thinking that the data i would use in the table would only be the number of right answers, for each question, from each group (consumer/company)?

    	Q1	Q2	Q3
    consumer
    employees

    post edited by tristram - 2008/05/12 18:13:42
    #6
    JK2
    Pro-Member
    • Total Posts : 14595
    • Reward points: 4963
    • Joined: 2003/07/14 15:07:10
    • Location: London United Kingdom
    • Status: offline
    RE: Stats help (JK2 possibly!) 2008/05/12 18:13:42 (permalink)
    Yes.

    Edit: Hang on.

    There are only two groups with only one condition?

    If there are only two groups - i.e. two independent populations, then the procedure you need is a Mann Whitney U test.
    post edited by JohnKerr2 - 2008/05/12 18:16:48

    #7
    tristram
    Pro-Member
    • Total Posts : 2746
    • Reward points: 1414
    • Joined: 2007/08/01 13:26:18
    • Status: offline
    RE: Stats help (JK2 possibly!) 2008/05/12 18:14:43 (permalink)

    ORIGINAL: JohnKerr2

    Yes.


    thanks. and just to double check, can i definatley use a chi square test for such a thing to prove the hypothesis?
    #8
    JK2
    Pro-Member
    • Total Posts : 14595
    • Reward points: 4963
    • Joined: 2003/07/14 15:07:10
    • Location: London United Kingdom
    • Status: offline
    RE: Stats help (JK2 possibly!) 2008/05/12 18:17:56 (permalink)
    See above.


    #9
    tristram
    Pro-Member
    • Total Posts : 2746
    • Reward points: 1414
    • Joined: 2007/08/01 13:26:18
    • Status: offline
    RE: Stats help (JK2 possibly!) 2008/05/12 18:17:57 (permalink)

    ORIGINAL: JohnKerr2

    Yes.

    Edit: Hang on.

    There are only two groups with only one condition?

    If there are only two groups - i.e. two independent populations, then the procedure you need is a Mann Whitney U test.


    there is two groups - consumer and company. and there is three questions which will be asked of both groups. and the correct answer to each question is no.
    #10
    JK2
    Pro-Member
    • Total Posts : 14595
    • Reward points: 4963
    • Joined: 2003/07/14 15:07:10
    • Location: London United Kingdom
    • Status: offline
    RE: Stats help (JK2 possibly!) 2008/05/12 18:19:46 (permalink)

    ORIGINAL: tristram


    ORIGINAL: JohnKerr2

    Yes.

    Edit: Hang on.

    There are only two groups with only one condition?

    If there are only two groups - i.e. two independent populations, then the procedure you need is a Mann Whitney U test.


    there is two groups - consumer and company. and there is three questions which will be asked of both groups. and the correct answer to each question is no.


    But your comparing numbers of correct answers between the two groups?

    (i.e. H-nought: The two sample populations score no differently
    H-a: The two sample populations score significantly differently)

    #11
    tristram
    Pro-Member
    • Total Posts : 2746
    • Reward points: 1414
    • Joined: 2007/08/01 13:26:18
    • Status: offline
    RE: Stats help (JK2 possibly!) 2008/05/12 18:20:36 (permalink)
    the null hypothesis is that the companies knowledge does not impact the consumers knowledge.
    #12
    Titch
    Pro-Member
    • Total Posts : 5515
    • Reward points: 5185
    • Joined: 2006/03/24 13:20:48
    • Status: offline
    RE: Stats help (JK2 possibly!) 2008/05/12 18:22:49 (permalink)
    This has 'Thread of the Year' written all over it!
    #13
    JK2
    Pro-Member
    • Total Posts : 14595
    • Reward points: 4963
    • Joined: 2003/07/14 15:07:10
    • Location: London United Kingdom
    • Status: offline
    RE: Stats help (JK2 possibly!) 2008/05/12 18:22:52 (permalink)

    ORIGINAL: tristram

    the null hypothesis is that the companies knowledge does not impact the consumers knowledge.


    I'm not sure then how your study will test this.

    Am I correct in thinking that you want to find out whether or not employees of a certain company know more stuff than consumers of products of that company?

    #14
    tristram
    Pro-Member
    • Total Posts : 2746
    • Reward points: 1414
    • Joined: 2007/08/01 13:26:18
    • Status: offline
    RE: Stats help (JK2 possibly!) 2008/05/12 18:26:39 (permalink)

    ORIGINAL: JohnKerr2

    I'm not sure then how your study will test this.

    Am I correct in thinking that you want to find out whether or not employees of a certain company know more stuff than consumers of products of that company?


    im not sure either, i hope there is a way though! its looking more likely i will need to rethink it. its testing employees not of the same company, on their knowledge of 3 consumers rights. and testing consumers on their knowledge of those 3 rights.
    with the idea of determining if the employees knowledge effects the consumers knowledge.
    #15
    JK2
    Pro-Member
    • Total Posts : 14595
    • Reward points: 4963
    • Joined: 2003/07/14 15:07:10
    • Location: London United Kingdom
    • Status: offline
    RE: Stats help (JK2 possibly!) 2008/05/12 18:29:27 (permalink)

    ORIGINAL: tristram


    ORIGINAL: JohnKerr2

    I'm not sure then how your study will test this.

    Am I correct in thinking that you want to find out whether or not employees of a certain company know more stuff than consumers of products of that company?


    im not sure either, i hope there is a way though! its looking more likely i will need to rethink it. its testing employees not of the same company, on their knowledge of 3 consumers rights. and testing consumers on their knowledge of those 3 rights.
    with the idea of determining if the employees knowledge effects the consumers knowledge.


    How would the employees' knowledge affect the consumers' knowledge?

    Have they met?

    #16
    tristram
    Pro-Member
    • Total Posts : 2746
    • Reward points: 1414
    • Joined: 2007/08/01 13:26:18
    • Status: offline
    RE: Stats help (JK2 possibly!) 2008/05/12 18:30:53 (permalink)

    ORIGINAL: JohnKerr2


    How would the employees' knowledge affect the consumers' knowledge?

    Have they met?


    good point. which i just thought of when actually thinking it through a bit more!
    i realised there must be some kind of relationship between the groups, and it isnt going to be possible to say there is.

    thanks anyway.
    #17
    JK2
    Pro-Member
    • Total Posts : 14595
    • Reward points: 4963
    • Joined: 2003/07/14 15:07:10
    • Location: London United Kingdom
    • Status: offline
    RE: Stats help (JK2 possibly!) 2008/05/12 18:32:50 (permalink)
    It might be worthwhile spending a bit of time drawing the whole study design out on paper - in a sort of tabular form.

    This can help make it clear which variables you're controlling and which you're hoping to give you the information you're after.

    It'll all become clear; trust me.

    #18
    Jump to:
    ©2017 All content is copyright of MuscleTalk.co.uk and its use elsewhere is prohibited.
    (posting guidelines | privacy | advertise | earnings disclaimer | contact us | supported by)
    © 2017 APG vNext Commercial Version 5.5