Confusion over split-plots?

P

======= Date Modified 01 42 2010 10:42:05 =======
Hi All,

Would someone please be able to clarify something that me and a collegue have been getting a bit confused over - what constitutes a split-plot design?!

So to use the same example that we were discussing - a researcher is interested in investigating the effect of food quantity and temperature on activity levels in mice. They allocate half of their mice to 'cold' labratories and half to 'hot' labratories. Within each lab, the mice are randomly allocated to one of three feed groups (small, medium or large feed).

Ok, so would this consitute a split-plot design OR do all of the mice have to take part in all 3 of the feed groups?

There seems to be different definitions of 'spilt-plot' depending on who you ask. So far I have heard the following definitions:

1. Subjects receive ALL levels of one factor but only ONE level of another. So in this instance, as the mice are only experiencing either HOT or COLD labs, then they would have to take part in all 3 feed conditions??

2. Two different experimental units. So in this case, we have lab as the unit for one factor (temperature) but cage as the unit for the other factor (food quanitity). So this definition does not say that all the mice must take part in each of the food groups?

Sorry if that is not very clear :$

Thanks in advance for any clarification!

C

Just spent 20 mins reading about split-plot designs and can't get my head around it! Gonna blame the fact I'm off sick today and brain not working! I can't really see why either of your examples are not factorial designs rather than split plot. Hopefully someone else will come along and answer this one.

C

OK. Think i've made sense of it.....to be a factorial design you'd need to randomly change the lab temp from cold to hot for each run or the experiement, which would be ridiculous. Instead it's a split-plot desgn as you split into the two separate labs the run the feed size experiment.

Nothing I've read up on suggests that the mice have to do all three levels of the feed experiment (most things talk about random allocation to one of the conditions) so I'll go with option 2 as correct.

Now if only I could bring myself to look at my own PhD work instead of daytime TV!

P

======= Date Modified 01 Nov 2010 12:38:32 =======
======= Date Modified 01 Nov 2010 12:37:56 =======

Quote From catalinbond:

OK. Think i've made sense of it.....to be a factorial design you'd need to randomly change the lab temp from cold to hot for each run or the experiement, which would be ridiculous. Instead it's a split-plot desgn as you split into the two separate labs the run the feed size experiment.

Nothing I've read up on suggests that the mice have to do all three levels of the feed experiment (most things talk about random allocation to one of the conditions) so I'll go with option 2 as correct.

Now if only I could bring myself to look at my own PhD work instead of daytime TV!


Yes, thats exactly right - its a split plot as we are working on the basis that it is not possible to randomise the temperature of the labs (technically, it could be done but for this example we're saying it can't). I agree with you, I didn't think the mice had to experience all 3 levels of the feed experiment but my collegue thought otherwise...

Thank you for taking the time to get your head round it & offer your opinion. It's nice to know I'm not the only one who finds it a bit confusing!

Good luck getting motivated with your PhD work, though if you are off sick I think you have a very good reason to watch daytime TV instead ;-)

16565