Video Tutorial: MPlus Data Setup

Univariate and Multivariate Outliers

A univariate outlier is a data point that consists of an extreme value on one variable.  A multivariate outlier is a combination of unusual scores on at least two variables.  Both types of outliers can influence the outcome of statistical analyses.  Outliers exist for four reasons.  Incorrect data entry can cause data to contain extreme cases.  A second reason for outliers can be failure to indicate codes for missing values in a dataset.  Another possibility is that the case did not come from the intended sample.  And finally, the distribution of the sample for specific variables may have a more extreme distribution than normal. 

In many parametric statistics, univariate and multivariate outliers must be removed from the dataset.  When looking for univariate outliers for continuous variables, standardized values (z scores) can be used.  If the statistical analysis to be performed does not contain a grouping variable, such as linear regression, canonical correlation, or SEM among others, then the data set should be assessed for outliers as a whole.  If the analysis to be conducted does contain a grouping variable, such as MANOVA, ANOVA, ANCOVA, or logistic regression, among others, then data should be assessed for outliers separately within each group.  For continuous variables, univariate outliers can be considered standardized cases that are outside the absolute value of 3.29.  However, caution must be taken with extremely large sample sizes, as outliers are expected in these datasets.  Once univariate outliers have been removed from a dataset, multivariate outliers can be assessed for and removed.  

Multivariate outliers can be identified with the use of Mahalanobis distance, which is the distance of a data point from the calculated centroid of the other cases where the centroid is calculated as the intersection of the mean of the variables being assessed.  Each point is recognized as an X, Y combination and multivariate outliers lie a given distance from the other cases.  The distances are interpreted using a p < .001 and the corresponding χ² value with the degrees of freedom equal to the number of variables.  Multivariate outliers can also be recognized using leverage, discrepancy, and influence.  Leverage is related to Mahalanobis distance but is measured on a different scale so that the χ² distribution does not apply.  Large scores indicate the case if further out however may still lie on the same line. Discrepancy assesses the extent that the case is in line with the other cases.  Influence is determined by leverage and discrepancy and assesses changes in coefficients when cases are removed.  Cases > 1.00 are likely to be considered outliers.

Creating a fact sheet to obtain participation in a study

Potential study participants need to know a series of information to decide if they would like to participate in your research. Some things they will need to know will include dates that the survey can be completed, how to gain access to the survey, what the study is about, if participation in the study is voluntary or mandatory, how long the survey will take, if there are restrictions as to who can participate, and if participants will remain anonymous or confidential.

In order to take part in a survey, potential participants will need to know how to access your survey as well as when the survey can be accessed. The researcher should inform the reader using information such as: The survey will be available Monday, March 18, 2013 - Monday, April 22, 2013. To take part in the survey you will need internet access. The following link will direct you to the survey: www.takemysurvey.com/123456. Another option is that the researcher will email the potential participants the link to the survey.

Prior to participating in the survey, the candidates will want to know what the research is about. Let them know who you are and describe what the aim of the study is. Provide insight into what you hope to achieve with the data that is collected. Let the potential participant know if the survey contains material that may be sensitive. Also, inform them of any inclusion criteria they must meet to participate in the survey. Inform them if participation in the study is mandatory or voluntary.

It is also helpful to inform the participant of how long the survey will take and if there is more than one round of surveys to complete. Providing the potential participant with accurate estimates of how long the survey will take will possibly decrease the likelihood of participant drop out. If the researcher grossly underestimates how long the survey will take, the participants will be less likely to complete the entire survey. The researcher should also inform the participant if the information will be anonymous, confidential, or available.

Dissertation Consulting...Getting the Help You Need

I felt a need to blog about getting dissertation support.  A client came to me a year ago seeking dissertation help.  At the time, she was not ready for that support and she subsequently spent another two (unnecessary) semesters at her university at $4,600/semester.  She then came back to me, with her financial aid maxed out at $160,000 and no further ahead in the dissertation process.  The point is this, sure, our services are a few thousand dollars, but we do help clarify your thinking, write data analysis plans and results chapters - and keep you moving.  

Get the help you need! Don’t wait until your financial aid is maxed-out and the thought of not finishing starts to take root. Statistics Solutions has been in business for decades, helping students from Walden, Phoenix, Capella, Nova Southeastern and many, many others.  Get the dissertation help and get out as soon as possible.  Our initial statistical consultation takes about 30 minutes and is always free at 877-437-8622 or Info@StatisticsSolutions.com

The Myth of Mixed Method Dissertations

I want dissertation students to be empowered in their dissertations, to learn something, and to graduate.  In the dissertation consulting business, Statistics Solutions has come across a lot of mixed method dissertations, and they’re seemingly on the rise.  When I ask the graduate student why they want to review two methodologies, then conduct and interpret both quantitative and qualitative methods, they invariably say because their advisor told them to.  The lament of the dissertation chairs are "one methodology is not enough" or I want you to "triangulate your findings."  There are literally tens of thousands of purely quantitative OR qualitative dissertations sitting in the library, so why this trend? 

First, I suspect that the chairs are getting pressure from the top, who are getting pressure from accreditation boards, to have more scholarly dissertations.  The answer to more scholarly dissertations is not more breath, but more depth.  Dissertations are to be original, add to the literature on a topic, and to provide a research experience to the student--this can certainly be done with one methodology.  And let's all be honest, most institutions are not research institutions--and maybe the chairs, administration, and accreditation boards should stop trying to put a square peg into a round hole, and appropriately focus on a dissertation that the students have been prepared to conduct.  (and by the way, just giving them a few dissertation classes is not preparation--the entire graduate experience is the preparation--if you can't do it, don't expect miracles in the last year).

Second, I suspect there is a mismatch between what the student’s comfort and preparation level and the chairs experience. There’s an old adage, that "your down on what you’re not up on."  Dissertation students should therefore pick wisely who their chair (and committee) is and simply ask them what kind of dissertations they have predominantly worked on, and even what the advisors own dissertation methods were like.    

Third, students are not empowered, nor have the experience, to say I want to do just quantitative or just qualitative.  I believe this happens for several reasons.  The dissertation is primarily an individual process and bouncing the process off others in a confidential manner with colleagues is difficult.  Second, there are a lot more people involved in the process: the chairperson, the committee, IRB or URR, the dean--so the chances of the dissertation getting stuck in one of these areas is much greater than just taking a class where the teacher has virtual total control of the process. I also think anxiety sets in.  Anxiety is natural since dissertation students are so close to finishing, the process is brand new, and their control seems to be at a minimum.

So what’s the solution?  First, pick the dissertation chairperson well and only once.  Changing chairpersons will almost always cost you more time and money.  You’re going to spend a year or more with this person in the dissertation process, so it’s worth it to have a few conversations prior to committing to that chair.  Second, as odd as it may sound, disengage emotionally from the process.  There are a lot of things you have no control over--sometimes the topic, department response time, and the structure of the process itself.  Just focus on what you can control, dispassionately, and consistently.  Finally, get support--social support, collegial support, statistics support, APA editing support--put the odds in your favor by getting help you need in the process with people who know you and the process.  The stakes are high, your time, energy, and monetary investment has been great, so finish strong, as quickly as possible, and get something out of the experience--that’s the kind of triangulation you truly need.  

Dissertation Chairs as Head Chefs

Dissertation chairpersons should be like head chefs—directing the doctorate student (the line cook) and the rest of the prep-chefs in the kitchen (i.e., the committee members).  But too often, the head chef defers his or her responsibility to the committee members (methodologists, statisticians members, or others).  This is very confusing for dissertation students as too often committee members want different things, in different orders, with different emphases.  Do you know what happens in a kitchen when prep-chefs have no leader?  You got it, the entire meal is uncoordinated and it doesn’t go together.  For dissertation students, this means frustration, more time in school, more tuition, and sitting on wonderful ingredients with no meal plan.  

While there was a lot of frustration in my own Ph.D. tenure, mostly because there were lots of dogmatic prep-chefs, when it came to the dissertation process, there was one, and only one, head chef—my dissertation chair.  Yes, the committee read the dissertation (probably the night before), but my chairperson had worked with me for over a year.  During that year we created, recreated, and corrected the methodology, the research design, and really the first three chapters were polished.  Consequently, my committee had very few comments because the head chef did his job!  In fact, during my dissertation proposal defense, it was more of a matter of explaining my research and defending it, and it went smoothly.  

So I implore all of those dissertation chairpersons to be head chefs.  Do not abrogate your responsibility to prep-chefs, even smart ones, for you put your doc student at a disadvantage, you lose their respect, your university loses credibility, and it’s just not right.  OK, you don’t have the statistical skill set, stick to qualitative dissertations, and if you have good quantitative skills than go promote yourself as a quantitative chair.  When you chair a dissertation, don’t just push them off to the methodologist, go get the skills yourself, if you do push them off make sure you are in the “cc” loop, and most importantly, get your dissertation student the unique support they both need and deserve that only you as a chairperson can provide.