ASA Connect

 View Only

  • 1.  split split plot design (SSPD)

    Posted 03-18-2019 11:53
    Dear all, greetings from Nigeria. Currently we are having arguement on SSPD. A colleague has an experiment on SSPD with fertilizer as mainplot treatment with two levels (NPK and Mucuna). Irrigation as subplot treatment with 3 levels (I1,I2,I3) and rate of fertilizer application as subsubplot treatment with four levels (r0=zero application, r2=2times before harvest, r4= 4times before harvest and r6=6times before harvest).therefore each sub sub plot is fertilizer x irrigation x rate of appliaction of fertilizer. So with first level of fertilizer(NPK), first level of irrigation (I1) and first level of rate of application(r0) we have NPK x I1 x r0. Can this be correct bearing in mind that r0 is zero fertilizer application.
    Best wishes all.

    ------------------------------
    Ikenna Nnabue
    Research Officer
    National Root Crops Research Institute.
    ------------------------------


  • 2.  RE: split split plot design (SSPD)

    Posted 03-19-2019 15:24
    Do you mean that NPK x I1 x r0 is actually the same as Mucuna x I1 x r0, because the r0 means that neither actually get the fertilizer? If so, it is possible to define your model so that these two estimated means are equal, but in practice, I've seen situations like this fit most often with a full factorial, which gives two separate means.

    ------------------------------
    Aaron Rendahl
    ------------------------------

    Original Message


  • 3.  RE: split split plot design (SSPD)

    Posted 03-19-2019 17:34
    There should be special consideration for  r0=zero application since no NPK is the same as no Mucuna.

    ------------------------------
    David Bristol
    Statistical Consulting Services, Inc.
    ------------------------------

    Original Message


  • 4.  RE: split split plot design (SSPD)

    Posted 03-20-2019 11:45
    One suggestion is to take the differences from the r0 subsubplot for the other 3 and convert to a 3 level design instead of 4, and proceed as you normally would for the analysis. You would have to keep in mind then that treatment effects represent the change in response relative to the zero level, which you're more or less treating as a control.

    ------------------------------
    Stan Altan
    ------------------------------

    Original Message


Related Content