10/16/2003 Quant
GET CAUGHT UP THIS WEEKEND! There's a paper to do next weekÉ
Influences on Correlations Coefficients
- Restriction
of range
- When
you restrict range it becomes come difficult to discern weak
relationships
- Example
Ð Stanford only has students with high SAT scores represented, so a study
done that looks at the correlation between grades and SAT scores using
only Stanford as your sample, you wouldn't see the whole range of
correlation.
- Use of
extreme groups
- can
lend impression of stronger correlation than if the more scattered center
were used
- Combining
groups
- Outliers
(extreme scores)
- Curvilinear
relationships
- Sample
size
- Reliability
of measures
Coefficient of Determination (r2)
- Coefficient
of Determination: the squared correlation coefficient
- The
proportion of variability in Y that can be explained (accounted for) by
knowing X
- Lies
between 0 and +1.00
- r2
will always be lower than r
- Percentage
of variance accounted for by the variable
- Often
converted to a percentage
Looking at Web Based Instruction Table 1. Biggest SD is for
final exam, so that probably shows that they had the most varied scores. SD
tend to be greater for Web course than FTF on all counts, so web-based scores
tend to have a greater range. Higher sample sizes for FTF.
Look at web-based study again and come back prepared to say
one thing about the Table 3 correlation table.