BookmarkSubscribeRSS Feed
DelBS
Calcite | Level 5

Hi,

 

I'm not confident in how to correctly go about specifying my modelin Proc Glimmix. This experiment was a RCBD with subsampling over time(3,27 hour time points). Below are my fixed and random effects.

 

with/without peptidesClass of peptidePeptideconcentrations (5 or 1 levels)time (2 levels)block (4 levels)
withCPPA6.25,12.5,25,50,1003,271,2,3,4
withCPPB6.25,12.5,25,50,1003,271,2,3,4
withCPPC6.25,12.5,25,50,1003,271,2,3,4
withCPPD6.25,12.5,25,50,1003,271,2,3,4
withCPPE6.25,12.5,25,50,1003,271,2,3,4
withCPPF6.25,12.5,25,50,1003,271,2,3,4
withnon-CPPControl peptide6.25,12.5,25,50,1003,271,2,3,4
without--03,27

1,2,3,4

 

The treatment without peptides is essentially a control to know the value of my response variables at a concentration of 0 - it applies to every peptide since the concentration of peptides is 0. Should I repeatedly enter this data for each peptide, since the values are applicable to every peptide? Or is there a way to enter the data once per experimental unit that was measured such that it is applied as concentration level of 0 for each peptide? Does it make sense for 'with/without peptides' to be a variable?

 

Another control in this experient is the control peptide. This peptide is of a different class than my other peptides. Correspondingly, I expect it to cause significant interactions as there is a different response with the control peptide. How do I include my data for the control peptide without compromising my ability to also analyse the significance of main effects and interactions within only the CPP class of peptides?

 

Are any of my effects nested here? For example, 'class of peptides' nested in 'with/without peptides', or 'peptide' nested in 'class of peptides', and 'concentrations' nested in 'with/without' peptides?

 


Any help would be appreciated.

 

Thanks

 

1 REPLY 1
sld
Rhodochrosite | Level 12 sld
Rhodochrosite | Level 12

The treatment structure of this experiment does not fit tidily into common ANOVA designs. I'll speculate below, but you are more likely to get useful responses from the community if you provide adequate detail about your experimental protocol, something like a Methods section. 

 

You have 35 combinations of peptides (7: A, B, C, D, E, F, Control) x concentrations (5: 6.25,12.5,25,50,100) plus one non-peptide control (a full factorial plus control; Google it). I'm going to assume that time is a repeated measure on whatever your experimental unit is, presumably, a unit within a block.

 

I don't see that any of your fixed effects are nested within any other fixed effect.

 

What is comprises a block? Does each block contain all 36 treatment combinations? What is the role of the non-peptide control (in other words, intuitively, what sort of comparisons would you make between the non-peptide control and the other 35 treatments)?

 

There's not really enough to go on here, and your experimental design might be incompatible with some of these approaches, but I would consider the following. (I'm ignoring the time factor.)

 

(1) Fit a "means model" with 36 treatments and use contrast statements to test hypotheses of interest.

 

(2) Fit 7 x 5 factorial (peptide x concentration) treatment structure, dropping the non-peptide control.

 

(3) Use the non-peptide control as a "baseline" covariate in an analysis of covariance model with a 7 x 5 factorial treatment structure.

 

(4) Subtract the non-peptide control value from the value of each treatment combination (or divide the value of each treatment combination by the value of the non-peptide control) in each block, and use the result as the response in a model with a 7 x 5 factorial treatment structure. This approach makes strong assumptions about the relationship between the non-peptide control and the peptide treatments that you should be aware of before you implement it.

 

I sense that you are not experienced with determining either an appropriate structure for an input data set or deterrmining an appropriate statistical analysis. You likely need more assistance than this forum can offer. If there is someone at your company/institution that can help you, seek them out! Do not overestimate what you know and underestimate the value of statistical consultation.

 

An extremely useful resource is

https://www.sas.com/store/books/categories/usage-and-reference/sas-for-mixed-models-second-edition/p...

Time spent reading this text is not time wasted.

 

Good luck!

sas-innovate-2024.png

Join us for SAS Innovate April 16-19 at the Aria in Las Vegas. Bring the team and save big with our group pricing for a limited time only.

Pre-conference courses and tutorials are filling up fast and are always a sellout. Register today to reserve your seat.

 

Register now!

What is ANOVA?

ANOVA, or Analysis Of Variance, is used to compare the averages or means of two or more populations to better understand how they differ. Watch this tutorial for more.

Find more tutorials on the SAS Users YouTube channel.

Discussion stats
  • 1 reply
  • 1616 views
  • 1 like
  • 2 in conversation