What is the best means in which to determine the sampling interval for 40 samples?

Question

I am interested in running an experiment in which 40 samples will be taken. If the samples are processed over a period of two weeks, how many days and samples per day should be tested? How did you come to this conclusion? Thank you.

Answer 1

In a real life situation, you would look at the items being sampled, and try to decide if the day of the week or the time of day was important to the thing being sampled.  

 

Water quality in a river downstream from a city, for example, is going to vary depending on time of day and day of week.  Water temperature, on the other hand, will vary mostly by time of day.    

 

If you were talking about an industrial process, then you might expect that it would change across the length of a worker's shift (does he make more mistakes when he's tired at the end of the day?)

 

If there's no special reason to do otherwise, you would probably just space out the 40 samples across the 14 days, which comes out to one every 8.4 hours or so (14*24/40 = 8.4).  That's assuming you are taking the first sample at the end of the first time period.  If you start at the beginning, then you only have 39 time periods, and they're about 8.6 hours each. 

 

You might choose to speed it up to one every 8 hours, taking them over 8*40 = 320 hours, which is 13 days.  That makes the actual collection schedule for each sample consistent, which is convenient for scheduling.

 

On the other hand, if getting statistically random samples was important, then you could also decide that you would randomly take each sample SOMEWHERE during the 8-hour time period, and then determine the exact time in the period to take the sample by rolling a die or some other random process which made each possible time equally likely.