November 11, 2009
>Sample size is also important for statistical analysis
>conventional wisdom N>30
>sample not>50% of population
>try to select enough to account for attrition of subjects
.Attrition means people have exercised their right to quit at any time
that they want.
???How to make sample representative of the population??
-every subject has an equal chance of being selected AND the selection of
one does not bias the chance of others.
.Imagine Lotto 649 the moment they pick #6 you can't pick 7, 8, and 9
.If you pick every 10th person it influences the chance of the others
-simple random selection - (Pull from "Hat", or the table of random
.You want to make sure that when you put names in a hat they are all the
.This would be best if you used your student #'s as they are all the same
.With replacement - means name goes back into the hat.
Page 1 of5 .Without replacement - means you name stays out of the hat.
.When you are in a raffle like the Princess Margaret - you hope it is
-Stratified random selection=divide population into the various subgroups
based on a characteristic or trait (gender, religion)
.Example out of 100, 70 are female and 30 are male. S=10
.Take 70 females and put all their names in a hat and select out 7.
.Select 30 males in a hat and pick out 3 of them.
.He will have 7 out of the ten being female and 3 out of 10 being male.
.Any time that you have a characteristic that makes the group different
than this type of sampling may be the route to go.
>Systematic sampling - subjects are selected using a "system"
>Cluster sampling - useful if the researcher has little knowledge of
characteristics and the population is widely scattered
.Suppose you were taking the names of schools in the GTA - there are a
lot of schools that start with Saint..... for Catholic schools sometimes
systematic sampling gives you a representative sample and other times it
Cluster Sampling - sampling schools in the GTA these schools are spread
out over a massive area.
.He would put all the schools in a hat and randomly select approx 20
schools and sample every teacher in each of those 20 schools.
Multi-stage sampling - is one more level to cluster sampling.
.There are two randoms one to pick the school and random to pick the
teachers in that school that are selected.
Can you have random sampling without replacement?
Example 1 out of 99 - the odds have changed instead of 1 out of 100.
.If you use replacement to fix the odds - 1 out of 100 put the name back
in the hat, then he pulls the name out it and if it is the same name he
puts it back and shakes it up again.
.Generally speaking we don't worry about that.
n is greater than 30 - n is the number of subjects.
.If n is greater than 30 for any experiment is that acceptable? No it
might be each condition might have n out of 30.
>Deliberate sampling - selects subjects with specific characteristics.
.Professor is interested in what the max VO2 of the elite cross country
.He would like to pick those individuals that are cross-country skiers.
.When you do deliberate sampling you cannot generalize to other areas.
Page 2 of5 .The results of biased because of the nature of your sample you started
>An experiment has two purposes
1. To help answer a research question
.An experiment helps to eliminate alternate.... (Professor changed
2. To control for possible alternate explanations.
>Success of study will depend on
-random assignment of subjects to groups
-ability to manipulate independent variable
-ability to control extraneous variables
-the timing of the measurements of the dependent variable
.Controlling extraneous variables with human subjects this is difficult.
.With animals - rats, mice it is easy because you have them in a cage and
hear is the food and hear is the water etc.
-Timing- when do you take measurements
.if you only measure them at the sa