Response to Another Student Discussion

DIRECTIONS: PLEASE READ CAREFULLY THE BELOW INSTRUCTION FOR THIS DISCUSSION:

Respond to this learner by identifying an element of the learner's posting that applies to your own understanding of logistic regression or your own research plans, AND ELABORATE ON THE POINT WITH SPECIFIC CONCERNS OR EXAMPLES.

Robert Laukaitis

Top of Form

U06D1 – Learned about Logistic Regression

In response to the question regarding what was learned about logistic regression, the discovery of different approaches appropriate for different types of research and data collected was most interesting. Depending on the methods by which data is entered into SPSS (IBM, 2016), there may or not be a theoretical basis for data entry or calculation. Contributions of each variable and the relationship to other predictor variables tends to drive a researcher in a particular logistic regression scenario. Sequential logistic regression seemed similar to the method by which participants respond to instruments. In the development, instruments tend to be validated by participant response in the order the questions were asked (Fowler, 2014). However, the stepwise logistic regression model assumes no underlying theory (Warner, 2013). This approach seems more exploratory, similar to the exploratory factor analysis (Flora, LaBrish, & Chalmers, 2012). The primary difference seems to be in the dichotomous nature of the data collected. Meeting specific criteria by which any particular approach might be impractical, yet might help the novice researcher in determining the approach that best describes relationships within a given dataset.

That said, the amount of data needed to effectively use the logistic regression might preclude a researcher from using the approach (Field, 2012). As with any other statistical approach, the logistic regression is subject to error if the assumptions of the approach are not met. The dichotomous outcome variables, independent variable outcomes, exhaustive and exclusive variable categories and the correct statement of the model used (Warner, 2013) suggest that Type II errors could be reduced with the appropriate amount of data and the correct model of logistic regression. Logistic regression will not be employed in my current research as the variables in the study are not dichotomous variables and therefore do not meet the assumptions of the approach.

Lastly, I found it interesting that the development of the approaches occurred in much the same way as the step-wise logistic regression was described. It would seem that in search of a good model fit, when existing models fail to describe the underlying model of the data, efforts continue in order to find methods to improve description and analysis of data that ultimately create better fits for data being analyzed. This can only help researchers find new ways to analyze new data and possibly review previous studies in search of deeper meaning.

References

Field, A. (2012). Discovering statistics. Retrieved Aug 12, 2016, from statisticshell.com: http://www.statisticshell.com/docs/repeatedmeasures.pdf

Flora, D. B., LaBrish, C., & Chalmers, R. P. (2012). Old and new ideas for data screening and assumption testing for exploratory and confirmatory factor analysis. Frontiers in Psychology, 3(55), 1-21. doi:10.3389/fpsyg.2012.00055

Fowler, F. J. (2014). Survey research methods (5th ed.). [VitalSource Bookshelf version]. SAGE. Retrieved from https://bookshelf.vitalsource.com/books/9781483323596

IBM. (2016). Statistical Package for the Social Sciences (SPSS) software. Retrieved Feb 28, 2016, from www.IBM.com: http://www-01.ibm.com/software/analytics/spss/

Warner, R. M. (2013). Applied statistics: From bivariate through multivariate techniques (2nd ed.). Thousand Oaks, CA: Sage. Retrieved from http://online.vitalsource.com/books/9781452268705

Reply Quote

Bottom of Form