I have looked up quite a number of research articles on this topic; none provide information on how the sample was determined. All these articles choose sample sizes that vary extremely from the other. This leaves me with no space for speculation. Can i get any help from an expert in this field of veterinary parasitology?
To make a good plan you need to know your results before you do the project. I have never been very good at this.
In an ideal world, you start by doing a preliminary project. What you really need to know is the existing infection rate. If the infection rate is less than 1% you will need a very large sample. You will also need a large sample size if the infection rate is 99% or more and it is important to quantify that last 1% (or less).
You need to also think about accuracy. Does it matter if the 95% confidence interval about a 30% prevalence is 4% to 73% or do you need a confidence interval of 29.5% to 30.05%? The latter will need a large sample size.
This sample size is for each treatment. If you want highly accurate results and prevalence is 0.004% then you will need thousands of birds from each treatment. Treatments can be planned (feed formula 1 versus formula 2) or unplanned (males versus females).
Facing an enormous sample size, and huge costs, consider pooling samples. If infection is based on a blood sample, then you can mix the blood from several birds and test the pooled sample for infection. Be careful about detection limits, but this can help you manage your expenses (cost/work time). There is a program for calculating prevalence and confidence interval from pooled samples called PooledInfRate: http://www.cdc.gov/westnile/resourcepages/mosqSurvSoft.html.
https://www.researchgate.net/publication/233576917_Rogers_Reexamining_the_Pooled_Sampling_Approach_for_Estimating_Prevalence_of_Infected_Insect_Vectors
Article Reexamining the Pooled Sampling Approach for Estimating Prev...
Wow! Thanks a lot, Dr Ebert. Your answer has begun to shed light on this issue for me.
Received with many thanks.
I agree totally with Dr Ebert for detection of "rare" events.
On the other hand, if you work on common chicken parasites like Eimeria sp. your problem will probably to have a picture of the distribution of the intensity of infection (oocyte counts in faeces, for ex) rather than an estimation of the prevalence (often close to 100%). In such case sampling 40-50 chicken at a given time would probably allow a rather good estimation of the skewness / curtosis of the data and then guide the choice of suitable statistical methods.
A typo. It should be kurtosis not curtosis, though I can find a few web pages that have used the latter spelling.
Just to clarify: With infection rates close to 100% pooling is not an option. Pooling does not usually make much sense with infection rates above 30%. Pooling is best with very rare events (
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size