Experimental Designs by unknow

Author:unknow
Language: eng
Format: epub
Publisher: SAGE Publications
Published: 2021-05-05T12:29:23.838916+00:00

Common tests of significance

Campbell and Stanley’s (1963) view is that when the unit of analysis is a single person or case, then the two most acceptable statistical tests are to compute gain scores or to use tests such as analysis of covariance (ANCOVA). In the gain scores approach, scholars first compute a pre-test–post-test gain score for each unit (e.g. T2 minus T1), and then compute a t-test value between experimental and control groups on these gain scores. However, ANCOVAs with pre-test scores as the covariance are ‘usually preferable to simple gain-score comparisons’ (p. 23). When the unit of analysis is an intact group or cluster, then the group means (rather than the individuals’ scores) should be used instead (e.g. Langley et al., 2020).

We note, however, that these common tests need to be applied when there is a logical assumption of representativeness, otherwise the fundamental purpose of conducting parametric statistical tests of significance such as the t-test or ANCOVA will not be met (Weisburd & Britt, 2014). In practical terms, unless the researcher has conducted probability sampling, or utilised the entire population of eligible cases, the assumption of representativeness will not be met. Using volunteers, cohorts of fixed populations or paid recruits has intrinsic value when the purpose is to generalise onto these very confined universes, to these populations only.

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.