Comparing Means Part 1
Part One – significance testing and the independent samples t-test
Comparing means
A step beyond comparing descriptives
In experimental research, the interest is about manipulating an independent variable
in order to make a causal inference
E.g. does listening to music improve exam performance – so we are comparing the
exam performance of music vs non music groups
Statistical significance
The technical question asked: how likely is it that we should have obtained the
results that we did, assuming that the null hypothesis about the population is true?
In other words – are differences between conditions so different we can truly say the
independent variable influences the dependent variable?
This is why we conduct tests of significance
Significance testing
Tests the probability of finding the differences in our results, assuming that the null
hyp. Of our population should be true
In psych, we are looking for a probability value (/p value) of 5% or less (i.e. p_<.05)
If significant results were “unlikely to occur by chance”
If not say results are non-significant (NOT insignificant)
^^as other factors could have influenced results hence its appropriate to say its non
significant
The T-test
Created by W.S Gosset under the pseudonym ‘Student’
o Testing whether yeast in beer batches affect quality
As gusset went under the name ‘student’ the test statistic is called Student’s t
When your performing statistical analyses and you get outputs you may find that
statistical tests have test statistic values such as students t e.g. pearson’s R or
spearman’s Rho
t = observed difference between - expected difference between population
sample means means (assuming H0 is true)
estimate to the standard error of the difference between the
two sample means
use this to create significant values but majority of time use SPSS to calculate t
values and p values also
Part One – significance testing and the independent samples t-test
Comparing means
A step beyond comparing descriptives
In experimental research, the interest is about manipulating an independent variable
in order to make a causal inference
E.g. does listening to music improve exam performance – so we are comparing the
exam performance of music vs non music groups
Statistical significance
The technical question asked: how likely is it that we should have obtained the
results that we did, assuming that the null hypothesis about the population is true?
In other words – are differences between conditions so different we can truly say the
independent variable influences the dependent variable?
This is why we conduct tests of significance
Significance testing
Tests the probability of finding the differences in our results, assuming that the null
hyp. Of our population should be true
In psych, we are looking for a probability value (/p value) of 5% or less (i.e. p_<.05)
If significant results were “unlikely to occur by chance”
If not say results are non-significant (NOT insignificant)
^^as other factors could have influenced results hence its appropriate to say its non
significant
The T-test
Created by W.S Gosset under the pseudonym ‘Student’
o Testing whether yeast in beer batches affect quality
As gusset went under the name ‘student’ the test statistic is called Student’s t
When your performing statistical analyses and you get outputs you may find that
statistical tests have test statistic values such as students t e.g. pearson’s R or
spearman’s Rho
t = observed difference between - expected difference between population
sample means means (assuming H0 is true)
estimate to the standard error of the difference between the
two sample means
use this to create significant values but majority of time use SPSS to calculate t
values and p values also
