Statistics

S_S116

Hi All,

I've been collecting data at four time points for this experiment I'm doing but at each time point I'm measuring a different sample (due to logistics I can't measure the exact same sample). I was going to use repeated measure ANOVA for the stats but now I'm not so sure considering the experimental unit (the sample I;m collecting the data from) is not repeated.

Any suggestions?

Ostrom

Can you elaborate on the research design? Is there some specific reason for the repeated measures and different units (i.e. do they differ on some important condition, or is the time of measurement important)?

If logistics are the only reason you collected from different groups, you could possibly pool the data depending on the sampling procedures used at each site. I suspect you may need something a bit more elaborate however. Generally, repeated measures tests are used for examining within-group differences.

S_S116

I was collecting measurement of ovarian structures from dairy cows on day 18, 19, 20 and 21 after they had been inseminated.


The question was weather the structures measured were different on any of the days measured between cows who were pregnant compared with cows who were not pregnant.
We also wanted to know did these structures change across time D18-21 after AI in either pregnant or non-pregnant animals?

The main question was if there are differences on any of the days between the two groups.

Black Swan

Were each of the samples random? Sample sizes?

Ostrom

S_S116 wrote: »

I was collecting measurement of ovarian structures from dairy cows on day 18, 19, 20 and 21 after they had been inseminated.


The question was weather the structures measured were different on any of the days measured between cows who were pregnant compared with cows who were not pregnant.
We also wanted to know did these structures change across time D18-21 after AI in either pregnant or non-pregnant animals?

The main question was if there are differences on any of the days between the two groups.

I'll have to avoid this one, as I don't want to steer you wrong. I don't work with experimental data, so anything I could advise would be from my own experience with different data structures. It sounds like a common enough design however - are there any papers in the field with details of analysis?

Do you have other covariates to include in the model? I'm thinking ANCOVA, which would control out the covariates and perhaps get you closer to the true time effect - although my experience in time series regression tells me there must be a more complex way to account for time in an experimental context (in ANCOVA, the dependent would be the structure, and the categories would be set by the time of collection). I'm only thinking this because of the sample group independence, but please do check with someone more experienced!

440Hz

S_S116 wrote: »

The question was weather the structures measured were different on any of the days measured between cows who were pregnant compared with cows who were not pregnant.
We also wanted to know did these structures change across time D18-21 after AI in either pregnant or non-pregnant animals?

The main question was if there are differences on any of the days between the two groups.

If you have one variable, the measured structure, and you want to test whether or not this variable varies from one group (i.e. pregnant cows) to another group (not pregnant cows) then a One-Way Between Groups ANOVA is your man!

The reason for doing an ANOVA is to see if there is any difference between groups on some variable.


Types of ANOVA

One-way between groups is used when you are looking at the differences between the groups. This is the simplest version of ANOVA.

A one way repeated measures ANOVA is used when you have a single group on which you have measured something a few times.

A two-way between groups ANOVA is used to look at complex groupings.

Hope that helps.

Kevster

It's sounding like there are 3 variables/dimensions in your dataset:

1, Time-point
2, 'Structure'
3, Pregnant or not pregnant

I would do a two-way ANOVA for each of the 'structures' measured. I assume they are simple haematological measures like blood pressure and platelet count?

So, in each two-way ANOVA, you'd be using time-point and pregnant Vs not pregnant. Once you complete a separate two-way ANOVA for each of your structures, you can see which one brings back significant values and at which time-point.

I'm unsure whether you need a parametric or non-parametric test. If unsure, always side on using the non-parametric equivalent, in which case your two-way ANOVA here would officially be called a Friedman Test for you.

Kevin