Statistical Assumptions Of Substantive Analyses Across The General Linear Model

1155 Words5 Pages
Nimon, K. (2012). Statistical Assumptions of Substantive Analyses Across the General Linear Model: A Mini-Review. Frontiers in Psychology Front. Psychology, 3(322). When using inferential statistics depends on the sampling technique used as well as the characteristics of the population data. This dependency does assume that the sample and population meet certain criteria. The criteria is called statistical assumptions, and if there are violations of these statistical assumptions that are not addressed, than the data may not be interpreted correctly. In particular, Type I or Type II errors may be increased or decreased incorrectly. Nimon’s (2012) article focused on assumptions associated with substantive statistical analyses across the…show more content…
Using unreliable data may cause underestimation which increase the probability of Type II errors. If there is correlated error, we may see unreliable data to be overestimated which would increase the risk of Type I errors. To satisfy the assumption of error-free data, researchers may conduct and report analyses using latent variables as opposed to observed variables. This technique is called structural equation modeling (SEM). This is where latent variables are formed from item scores, the former of which becomes the unit analyzed. It is important to note that SEM is a large sample technique. A researcher can also delete a few items to raise the reliability of an observed score, but this should be reported along with the accompanied estimated of the reliability with and without the deleted items. Some researchers consider measurement to be an assumption while others do not. Nimon (2012) does state because measurement levels play such a pivotal role in statistical analyses decision trees, it would not be recommended to lower the measurement level of data unless certain characteristics are met. This may discard important information and may produce misleading or erroneous information. Yet another assumption, is the assumption that inferential statistics in psychological and educational research produce population data that is normally distributed. This does depend on the analysis conducted, such as univariate, multivariate,

More about Statistical Assumptions Of Substantive Analyses Across The General Linear Model

Open Document