Data Screening

Data Screening

Once data from a research study is gathered and has been entered into SPSS, researchers must examine their data to be sure they can validly interpret their results. Valid interpretation of data is reliant on two data features:

1. The data must meet the assumptions of the analysis procedure.

2. The data in the data file are "an accurate representation or transcription of what was provided by research participants as their original responses or what was provided by archival sources as original data" (Meyers, Gamst, & Guarino, 2017, p. 31).

Contribution by: Britany Kuslis, WCSU Cohort 8

Reference:

Meyers, L., Gamst, G., & Guarino, A.J. (2017). Applied multivariate research: Design and interpretation. Thousand Oaks, CA: Sage Publications.

---

Data Cleaning

This process of representing original data. In its initial phases, involves the researcher looking for incomplete data that may skew futher data screening. For example if there is a 5 question subscale where all the scores are added and a mean derived, if a respondent does not answer a question, that would skew the subscale, and that set of answers should be removed.

Contribution by: Mykal Kuslis, WCSU Cohort 8

Reference:

Meyers, S., Gamst, G., & Guarino, A.J. (2017). Applied multivariate research: Design and interpretation. Thousand Oaks, CA: Sage Publications. (p. 31-33)

Value Cleaning

Value cleaning is ensuring the values are "within the limits of reasonable expectation" within the "to the extent that it is possible...within the bounds of feasibility"(Meyers, Gamst, & Guarino, 2017, p. 32). For example, you want to ensure the age of a presumed adult is not 9 years old or that a response to an item rated on a likert scale of 1-5 is not a 6 or an otherwise value that is not within the bounds of the study.

Contribution by: Britany Kuslis, WCSU Cohort 8

Reference: Meyers, S., Gamst, G, & Guarino, A.J. (2017). Applied multivariate research: Design and interpretation. Thousand Oaks, CA: Sage Publications.

Outliers

Outliers are values that are "extreme or unusual values on a single variable (univariate) or on a combination of variables (multivariate)" (Meyers, Gamst, & Guarino, 2017, p. 48).

The presence of outliers can greatly impact the results of an analysis for two major reasons:

(1) The mean of the variable might no longer be a good variable and

(2) Outliers will yield a difference that when squared will produce a value too large that will skew the computation.

Outliers may signal "anomalies within the data" that will likely need to be addressed prior to moving forward with the statistical analysis (Meyers, Gamst, & Guarino, 2017, p. 48).

Contribution by: Britany Kuslis, WCSU Cohort 8

Reference:

Meyers, S., Gamst, G., & Guarino, A.J. (2017). Applied multivariate research: Design and interpretation. Thousand Oaks, CA: Sage Publications.

Detection of Multivariate Outliers: Mahalanobis Distance

The Mahalanobis Distance statistic measures "the multivariate 'distance' between each case and the group multivariate mean (known as centroid) taking into account the correlations between the variables" (Meyers, Gamst, & Guarino, 2017, p. 52). This method is used to determine if there are scores that vary from the mean of a set of DV's. The Mahalanobis distances details how far a case is from the group center mass of the predictor or IV's. The greater the distance the higher the possibility of a multivariate outlier. According to Lawrence S. Meyers, Glenn Gamst and A.J. Guarino, "Each case is evaluated using the chi square distribution with a stringent alpha level of .001. Cases that reach this significance threshold can be considered multivariate outliers and possible candidates for elimination. This approach is also not without its critics (e.g., Wilcox, 2012) for alternative approaches to multivariate outlier detection" (Meyers, Gamst, & Guarino, 2017, p.53).

Identifying Multivariate Outliers with Mahalanobis Distance-->[1]

Mahalanobis Distance -->[2]

Contribution by: Britany Kuslis, WCSU Cohort 8

References:

Clapham, Matthew E. “Mahalanobis Distance.” YouTube, YouTube.com, 2016, www.youtube.com/watch?v=spNpfmWZBmg.

Grande, Dr. Todd. “Identifying Multivariate Outliers with Mahalanobis Distance.” YouTube, YouTube.com, 2016, www.youtube.com/watch?v=AXLAX6r5JgE.

Meyers, L., Gamst, G, & Guarino, A.J. (2017). Applied multivariate research: Design and interpretation. Thousand Oaks, CA: Sage Publications.