Journal article review
Journal article review
The title of the article is analysis of covariance (Ancova) by Linda Kudee. It is a 29 page long article that discusses the use of Ancova and how it can be used to control the power of the test of independent variables. The main issue highlighted by this article is that as Morrow A.J (2009) puts it, ANCOVA is an extension of analysis of variance in which additional variable is added called covariate to the equation. And covariate is added to control statistically for the effect of the variables, and that besides covariate increases the sensitivity of the statistical test (Morrow, 2009).
In the article the researchers are interested to see the effect of two different therapies (stalking-cruel-to-be kind therapy and psychodynamic therapy) on stalking behavior. The independent variables are the therapy approaches, and the dependent variable is the stalking behavior after therapy (Posttest). The co-variable is Pretest stalking behavior (Morrow, 2009) but the concern here is that in a research comparing various groups of participants, observed differences before the test are mainly due to chance rather than being related in any way to the group variable thus the difference that are seen in the two groups due to various therapies may have nothing to do with the pretest stalking behavior.
As Wholey (1994) puts it the outcome of the two therapies are likely to differ because of the preselect ion differences that existed. This means that the validity of the results that are obtained after the test maybe questionable.
In my view the article begins on wrong note by stating in the first paragraph that Ancova is used to statistically control the effect of variable but I believe that control is a wrong word to use as it relates to Ancova. It is important to point out here that some scholars and researchers think that the primary role of ‘Ancova is to control or to remove’, some parts of the covariate but this not so. Ancova can help in randomly selected groups and this can produce very spurious results
In this article there is an attempt to control the difference that are likely to occur because of the two approaches of therapy that have been used and their outcomes but this is just very inappropriate.
It makes no sense for researcher to ask how the stalking behavior of someone undergoing psychodynamic therapy would be when cruel-to-be kind therapy if applied. Ancova may, in way be able to control this. Due to the specification that may be brought by covariate, differences remaining intact and the outcome of the behavior after treatment may be biased.
Again we are not told how the groups (two different therapies) would differ if they did not differ in the covariate and what and what else the researcher can do without analysis of the covariates.
Another great thing that is worth mentioning is that the article doesn’t have an abstract that gives an overview of the journal is about. Having an abstract would be a great way of endearing people to the article and raising their appetite and willingness to read the whole article. The many table and histograms that have been used would also need lots of technical skill and perhaps know-how to understand them. The article dwells so much on the assumptions that are expected when doing analysis of covariance. Though it is important to mention the assumptions but much more time would have been taken in empathizing the main ideas of the article.
The article has come short especially as it relates to citation and referencing because one of the biggest sources and perhaps the most cited(Morrow,2009) has not be included in the reference section and that is strange. I would have expected that not to be the case. Not just that I feel that article that have sufficient academic grounding. In terms of having other source that have been peer reviewed. For a paper such as this one it would be important to do serious research and credible sources.
I felt that article had too many tables, figures and calculations and that should have been backed by equally enough explanations. The problem with figures is that they have a way of making a document come across as academic and well-researched but in most cases that is not what you get. The article hasn’t done as much to explore the impact of the covariate in a single outcome. The pretest stalking behavior hasn’t been clearly outlined and explained. Instead what is there the number of hours that are pent stalking and how they differ and that also depends on the kind of therapy that has been used. The researcher is investigating the relationship between the two therapies and how they might impact the post therapy behavior but this suggested relationship could be strengthened by application of covariates. They might consider looking at how insecurity low-self- esteem .We might find that insecurity has something to do with stalking behavior and this does not differ regardless of the therapy method that is being applied. In that case, we might insecurity as covariate and expect to find out that the relationship between stalking and low-self-esteem is strong.
The article also delves into the assumptions that are associated with the analysis of Ancova as outlined by Dr Morrow. The first assumption is that of existence of outliers. An outlier is an observation is far and perhaps different form other observation points. This could be caused by variations in measurements or experimental errors and it is important to mention that it can greatly impact the result of an analysis.
Another assumption that has been highlighted in the article is the homogeneity of regression which means that there is a constant relationship between the covariates and the outcome. That is to say that the relationship between the outcome and the covariate is unchanging. The article states the assumption that there is normality of independent variables (Field, 2013).This kind of assumption opens the door for vulnerability of correlated scores which leads to Type 1 errors. It is hard to show if the distribution sample is normal unless you take multiples samples and form a sampling distribution
The other assumption of Ancova made in the article include the homogeneity of the variance which suggests that independent variables should have equal variances, and the covariates should have variances across independent variables and with the help of the graphs the researcher has attempted to compare physically the correlation outcomes. But much work needs to be done here to objectively examine by way of statistical analysis. We need to find out if the regression slopes differ between the two therapies by use of a custom model in SPPS. This is because calculations for Ancova can be very complex but with the help of Ancova we can be able to homogeneity regression for each one of the therapies.
Another one is multi-linearity which says that If covariate is more than one there should not be highly correlated with each other meaning the correlation in terms of absolute value should not be greater to an 0.8.This highlights an important point of variance partitioning and this something that is used and greatly abused especially in the field of social sciences. It makes it difficult to explain them individually when the covariate and in thus people can make some odd statements that end up confusing even further.
And lastly the assumption that running an analysis of covariance there is need to ascerta if there is any missing data because it can potentially greatly lower the power of analysis of covariance.It would have been good for the article to begin by ignoring the assumptions and going straight into the analysis which is the main aim of the article. As Wainer (1991) said that that the appropriateness of ANCOVA should not only depend on the statistical assumptions but on the nature of the question that has been posed. This article requires the analysis that may be effectively covered by statistics and that is how I thought it should have been handled from the beginning.
. As Chapman & Chapman (1973) found out there is no method that can address the question if the two groups that differ on variable A would differ on variable B. The most and perhaps the only legitimate way of the use of analysis of covariance is reduce the variability of scores in groups that were randomly selected and that randomly vary. The experimental methods of carrying out the analysis have not be adequately explained in the article. What we get instead is that there are differences but that hasn’t been well worked and developed for to be able to see and possibly duplicate.
In conclusion, it is my hope that my critique will help in portraying the proper use of Ancova. Researchers should also consider using other methods that can best help them in improving their design.
Chapman,L,J & Chapman,L,P (1973).Disordered thoughts in schizophrenia. New York: Appleton-Century-Crofts.
Field, A. (2013). Discovering Statistics using IBM SPSS. SAGE Publications Ltd
Morrow, K. (2009). Mental health of college students. New York: Nova Science.
Wholey, J. (1994). Handbook of practical program evaluation. San Francisco: Jossey-Bass.
Place an Order