Programming the statistical procedures from SAS

T-Tests vs Other Modeling types

Reply
Occasional Contributor
Posts: 5

T-Tests vs Other Modeling types

Hi there,

 

The data that I am working with is egg production data over a course of 12 years. I was thinking of just taking the average flock production from the first and last year and putting it into a t-test to see if it was significantly different. My adviser thinks that there is probably some other type of modelling SAS can do for me that would provide me with more data. Do any of you good people happen know what that would be? (Broad question, I know).

 

I'm not very familiar with SAS at all and I haven't started working with it much yet. But I have to explain how I'm going to use SAS for a research proposal, which is a little hard when I don't entirely know how it works.

 

If anybody can help me get a better idea of what this modelling would look like compared to a T-test, that would be greatly appreciated!!

Respected Advisor
Posts: 4,606

Re: T-Tests vs Other Modeling types

Your scientific hypotheses should not depend primarily on which statistical tools are available. You should state the hypotheses first and then ask: Can I use SAS to test them?  Given the wide array of models that SAS can accomodate, the answer will depend mostly on your sampling plan and on the amount of data available.

PG
Occasional Contributor
Posts: 5

Re: T-Tests vs Other Modeling types

Hi PG,

 

Thanks for replying to me! My hypothesis isn't based on the stats available and that part is all done and sorted out.  It's the materials and methods part of the proposal that this is going in where I need to be able to better explain what stats I plan on using.

Grand Advisor
Posts: 16,893

Re: T-Tests vs Other Modeling types

How are you planning to test your hypothesis?
Occasional Contributor
Posts: 5

Re: T-Tests vs Other Modeling types

All the egg production data is collected at the farm on a daily basis for every hen.  It is stored in an  Excel file and the data for each bird is broken down by number of eggs per week (ideal number being 7 eggs).  I was going to take the files and chart the flock average for each breed of bird for every year to see if there were trends or spikes.  Then I was going to take each breed average from the first year and run it through a t-test for each breed average for the most recent year.

Respected Advisor
Posts: 4,606

Re: T-Tests vs Other Modeling types

WHAT is your hypothesis?

PG
Occasional Contributor
Posts: 5

Re: T-Tests vs Other Modeling types

The hypothesis is that since selection for the birds has not been based off of total egg production in the 12 generations we have had them, there will be no statisical difference between their total egg production from the year we first got the birds and now. 

Respected Advisor
Posts: 4,606

Re: T-Tests vs Other Modeling types

@Reeza was right, this is about chicken. At this point, I feel like calling in @Linlin (an expert in both SAS and hens)! How do you account/control for the bird's age? Is year tthe same as generation number?

PG
Occasional Contributor
Posts: 5

Re: T-Tests vs Other Modeling types

Yes, sorry, I should have explained that better, year and generation are basically interchangable in this case since we hatch out a new generation at the end of March every year and keep them around for a year before culling them to make space for the subsequent generation.  All the birds are hatched on the same day so all of them are the same age for each generation.

Respected Advisor
Posts: 4,606

Re: T-Tests vs Other Modeling types

[ Edited ]

I think you should be using models for count data (individual bird annual production). You could test for differences between years (generations) and/or an overall time trend. The most appropriate SAS procedure for the task is proc glimmix, which unfortunately is not easy to use. So you might want to secure some help before you commit to it.

If you have a licence to SAS/ETS, you could use proc countreg, which is dedicated to count data analysis and a bit easier to master.

PG
Grand Advisor
Posts: 16,893

Re: T-Tests vs Other Modeling types

Please, please get a statistician to help you. Someone invested money to raise birds for 12 years, hopefully they can see the benefit in hiring a statistician to ensure appropriate data analysis. This is not a straight forward analysis and it's probably worth getting the most out of the data.
Grand Advisor
Posts: 16,893

Re: T-Tests vs Other Modeling types

Hi @Justin_Allison, welcome to the SAS Forums.

 

If you'd like to post tutorial style posts, you can use an Article rather than create a thread or respond to a two year old thread.

Additionally, since you are new to the forums it would help if you post your links as full link rather than hyper link to first establish some trust. Many users will not click on a link where you cannot see the full address.

 

 

 

Grand Advisor
Posts: 16,893

Re: T-Tests vs Other Modeling types

SAS = Statistical Analysis Software. Determine your analysis plan and then how SAS can support it. It's only a tool, the methods behind your statistical analysis is more important.

We don't have quite enough information to help at the moment. It depends on how your data is collected and what it represents. Do you have egg production over time as an aggregate measure, or is it by chicken or coop or some other aggregate area. I would assume that not all chicken types would have the same production levels. Depending on the answers you may be looking at some sort of repeated measures analysis and/or some sort of time series analysis. If you have different levels of aggregation it could be a hierarchical model.
Ask a Question
Discussion stats
  • 12 replies
  • 428 views
  • 4 likes
  • 3 in conversation