Results 1 to 4 of 4

Thread: Pseudoreplication for my PhD Thesis: any advices?

  1. #1
    Newbie
    Joined
    Dec 2014
    From
    Italia
    Posts
    3

    Pseudoreplication for my PhD Thesis: any advices?

    Hello to everybody, that's my first post in the forum.

    I apologize for i) my bad english and ii) for the lenght of the post

    _______________________

    I have some doubts concerning the experimental design for my PhD thesis.

    As you can see from the picture attached I'm going to compare three different land management in three different areas: (i) plain, (ii) hilly, (iii) mountain trying to investigate the level of CO2 emission for each management option (pasture, wheat ...).

    I'm not a statistic expert but since soils are not comparable between plain, hills and mountain, I think I need to perform ANOVA inside each area and NOT between the areas.
    And I assume each field is infact a "treatment". So I'm going to perform pseudoreplication because wheat soil crop is a... subsample of wheat soil.


    By the way my main questions are:

    i) Was I enough clear?

    ii) As you can see from the picture, one grey square = to a repetition. So I have 3 repetition per field (treatment). How many (pseudo) repetition for do I need to make sure my analysis will be statistically strong enough?
    Attached Thumbnails Attached Thumbnails Pseudoreplication for my PhD Thesis: any advices?-situation.png  
    Follow Math Help Forum on Facebook and Google+

  2. #2
    MHF Contributor
    Joined
    Sep 2012
    From
    Australia
    Posts
    6,602
    Thanks
    1712

    Re: Pseudoreplication for my PhD Thesis: any advices?

    Hey Muzze.

    How are you partitioning the information within each section?

    The reason I ask is whether you have any reason to believe that one section can affect another - and remove this if it can. If the information has no significant correlation then you have independence and you can use something like an ANOVA to test a difference in any one of the means.

    Justifying the model requires justification of the physical attributes of your stochastic process and this requires an understanding of the domain knowledge in a deep manner.

    I think it's an idea to try and define a statistical model of sorts, along with the tests you want to do and then analyze the physical situation to see how it supports or doesn't support your choice of model, assumptions, tests, and choice of data structure.
    Follow Math Help Forum on Facebook and Google+

  3. #3
    Newbie
    Joined
    Dec 2014
    From
    Italia
    Posts
    3

    Re: Pseudoreplication for my PhD Thesis: any advices?

    Hi Chiro!

    I feel a little bit embarassed to ask this things that may appear a bit ordinary but my stat. course starts in April. However since i'm the II° year of PhD I need to collect info as soon as possible. I'm very glad you aswered me.

    I'll try to put the issue in a easier way:

    Let's say I have 3 field, each field has the same crop, exactly the same. Now, how many sampling points do I need for each field in order to have good data?

    Let's say that the answear is 30 sampling point per flied --> can I use ANOVA to test a difference between them?

    Thank you againPseudoreplication for my PhD Thesis: any advices?-dd.png
    Follow Math Help Forum on Facebook and Google+

  4. #4
    MHF Contributor
    Joined
    Sep 2012
    From
    Australia
    Posts
    6,602
    Thanks
    1712

    Re: Pseudoreplication for my PhD Thesis: any advices?

    That depends on your level of confidence and how wide you want your interval to be.

    If you were say using a normal distribution for your estimator (you may have to use something else but bear with me for the moment) and you wanted say your interval to be 2I units wide then you are solving for:

    P(|X| < I) = 1 - alpha where alpha is the significance level (for 95% you have alpha = 0.05 and so on).

    Basically if you have a sample variance s^2 then you are solving for I given:

    s/sqrt(n) < I (we are not counting the mean yet - only interested in the length of the interval) which imples

    n > s^2/I where 2I is the total interval and I is the size either size of 0.

    So basically you need to find the value of n that matches this (the minimum whole number n) and this is used to pick your sample size to guarantee (under the assumptions of the model) that the error in some confidence level will be restricted to that sized interval.

    There are more general techniques that look at power calculations (which do the same sorts of things) but they beyond the scope of this post.

    Note that if you decrease the interval the n figure increases (as expected since more information is needed for smaller interval) and if I increases then you need less n.

    You will need to do the above for each crop individually you are to look at comparisons.
    Follow Math Help Forum on Facebook and Google+

Similar Math Help Forum Discussions

  1. Replies: 3
    Last Post: Mar 2nd 2012, 06:25 PM
  2. Church's thesis
    Posted in the Discrete Math Forum
    Replies: 1
    Last Post: Apr 1st 2011, 09:19 AM
  3. Ms thesis
    Posted in the Advanced Applied Math Forum
    Replies: 0
    Last Post: Jan 24th 2010, 07:13 AM
  4. my thesis
    Posted in the Pre-Calculus Forum
    Replies: 0
    Last Post: Jul 9th 2006, 03:19 AM

Search Tags


/mathhelpforum @mathhelpforum