1. ## Question about Repeated Measures and Type 1 Error

I'm working on analyzing a study where there are a number of buckets of soil and observations were recorded on how many plants are visibly growing from each bucket. The observations were recorded 18 times, over the course of 2 months (not at strictly regular intervals).

The thing is, there was a very noticeable change over time for the first 3 observations, but then they more or less flattened out - plants were not prone to dying and new plants were not prone to growing. So, basically, it's like we're observing the same plants repeatedly with some very minor variation.

A repeated measure (time) ANOVA on all the data gives a very significant result for the main effect of soil condition. A simple ANOVA with the observations limited to the final day of recording did not yield a significant main effect (not even close). The actual observed difference on the final observation was pretty close to most of the others.

This has me thinking... am I unreasonably increasing both my power and my type 1 error risk via inflating degrees of freedom by performing an ANOVA (with soil condition as predictor) that includes all these observations with time treated as a repeated measure?

If so, why is treating time as repeated measure inappropriate here? It makes sense to me intuitively why it would be, but justifying for or against has me puzzled.

Thanks.

2. ## Re: Question about Repeated Measures and Type 1 Error

I guess the issue is whether the repeated measures analysis with the focus on the main effect of condition is testing the hypothesis you're actually interested in. In terms of the main effect of condition, the null hypothesis it tests is that, within every time point, the true means are identical across the two conditions.

But perhaps your interest is really in whether the plants in one condition changed faster than those in another - in that case, it's the interaction between time and condition that you should be focusing on. And if you really only care where the plants end up at the end of the time period, then just a oneway ANOVA on the last observations might make sense.

3. ## Re: Question about Repeated Measures and Type 1 Error

In terms of the main effect of condition, the null hypothesis it tests is that, within every time point, the true means are identical across the two conditions.
At any moment? I thought the point of repeated measures was it tested if a value changed from one point in time to another?

4. ## Re: Question about Repeated Measures and Type 1 Error

Originally Posted by noetsi
At any moment? I thought the point of repeated measures was it tested if a value changed from one point in time to another?
Yep, but notice I was talking about the main effect of condition. There'll also be a main effect of time, and usually time*condition interaction term.

5. ## Re: Question about Repeated Measures and Type 1 Error

I hope your stats goes well

 Tweet

#### Posting Permissions

• You may not post new threads
• You may not post replies
• You may not post attachments
• You may not edit your posts