Is Quantum designed to automatically check for multiple comparisons / correct for false positives when stat testing?

Technote (troubleshooting)


Problem(Abstract)


Multiple comparisons corrections would occur within the Quantum
tables. If you are doing hundreds of mean tests using 95% and 90% as
your criteria, just by chance 5% to 10% will give you false positives.
There are corrections you can make to the stat test that take into
account the number of tests you are doing and correct the statistics
accordingly. Can Quantum make those corrections.

Resolving the problem

The only adjustment Quantum can make/differences between Quantum
formulae and text book statistical formulae are the following for the
stats tests applied using the keyword tstat:
1) Quantum makes
adjustments for when weighting is applied. This is reduce the
likelihood that any significant differences are due to the fact that
weighting has been
applied rather than due to true significant differences.
2) Quantum
will allow you to apply an overlap adjustment for cases where
respondents can appear in more than one column being tested, for
example where the columns are categories
from a multicoded question. If overlap were not
applied, the respondent's answers would be treated as if they were to
come from two independent respondents.
3) When applying the tstat prop
test you can add the option corr to make what is known as a continuity correction adjustment, This is because when you apply stats testing
the assumption is that the variables are distributed as a normal
continuous one but proportions are
distributed as a discrete binomial one, so the adjustment is made to
correct this difference.


Rate this page:

(0 users)Average rating

Add comments

Document information


More support for:

SPSS Data Collection Heritage
Quantum

Software version:

5.8

Operating system(s):

Platform Independent

Reference #:

1620621

Modified date:

2012-12-14

Translate my page

Machine Translation

Content navigation