Starting the Semester With A Little R

In our last blog post, we told you that in our Applied Methods in Social Work Research classes that we taught during the summer,  we introduced R on the first day of class EVEN though we don't get into data analysis until much further along.  

We thought that introducing students to R early would build confidence as we taught skills incrementally. Comments from two students from our last post provided support that this may be the way to go in the future, too.

In this post, we thought we would share a little bit about how we approached this.

First, the Applied Methods courses in our school are taught in computer labs, so each student had a computer available during class time. This made our lives infinitely easier when we did in-class activities.  

During the first class we did some really simple stuff:

1)  We showed the students how to open RStudio and gave them a tour of the panes.  We showed them that they would be most interested in looking at datasets, called data frames in R, in the Environment tab.  We showed them that they would enter commands and view output in the Console.  We also showed them that they would view graphs in the Plots tab in the lower right pane.

2)  We opened an existing dataset and showed the students the tabular view by clicking on the spreadsheet icon in the Environment tab.  This was a great way to first introduce students to the idea of multiple observations/cases/subjects and variable names.

3)  Finally, we encouraged students to download R and RStudio at home since in the following classes there was going to be homework that needed to be completed outside of class.  To provide additional support, we pointed them to two videos on our YouTube channel, SSDforR.  The first was on installing R and RStudio in a Windows environment and the other was on installing on a Mac.  These videos are also accessible from our website at:  https://www.ssdanalysis.com/Videos.html

We would love to hear how this worked for our students so feel free to post a comment or send us an email! Also, if you teach or train people to use R, we would also love to hear your ideas! 



 

Using R in the Classroom (or How We Tried to Inspire Reluctant Students)

The summer is nearly over and we are preparing for our fall classes.  As we are readying ourselves for a busy semester, we thought we would reflect back over the summer semester and let you know how we used R with our students taking Applied Methods in Social Work Research, a general course in  research methods.

This was the first semester where ALL of our Applied Methods sections used R instead of proprietary software, so we took this opportunity to make some small, but substantial changes to our teaching methods.

For the first time:

• We introduced R on the first day of the semester EVEN though we don't get into data analysis until much further along.  Why did we do this?  We thought that introducing students to R early would build confidence as we taught skills incrementally.
• We gave brief and simple homework assignments after most classes.  This is something we had been hesitant to do in the past, as not all students had access to statistical software off-campus (i.e., we never required students to purchase expensive propriety software previously).  Since R and RStudio are free, safe downloads, everyone could access it from their home computers.  Some authors have suggested that requiring students to do independent research work outside of the classroom  is related to more positive attitudes towards research and increased abilities (see Elliott, Choi, & Friedline, 2013).
• We had an open lab at least once a week where students could join us and ask questions about R or their homework assignments.  This gave students the opportunity to get a little one-on-one attention in areas where they thought they might need some individual help.
• The students completed a semester-long research class evaluating factors related to satisfaction with social work education.  They got involved in all aspects of the project from identifying research questions, creating hypotheses, constructing a survey, collecting data, to analyzing data and reporting findings.

By the way, if you are reading this blog and you were one of our students this summer, we welcome comments and constructive criticism, particularly on how R worked for you!

If you find this interesting, we invite you to join us at The Council on Social Work Education Annual Program Meeting in Tampa.  We are presenting a skills workshop on Saturday, October 25th at 11 am.  The title of the workshop is Using R to Build Research Skills and Research Capacity in Master's-Level Students.  As we get closer to APM, we will share with you a bit of what we will be presenting, but, as usual, we will be able to provide you with much more content and many more resources than makes sense in a blog.

Feel free to check out our website at www.ssdanalysis.com and, as always, we welcome your comments!

R, R and more R!

When we first set out to write this blog, we were really only thinking about using R to analyze data with single-subject designs - and SSD for R was born and our book was written!

But, as we worked on that software and book, we did a lot of reading and spoke to a lot of people.  We piloted SSD for R with students, and we talked to each other about it a lot.  We came to the same  (anecdotal) conclusion - using R is just as easy for students to use as any other statistical software package.  Even those really slick menu-driven packages.  Even those really expensive proprietary packages that EVERYONE uses.

In fact, R seemed to have a distinct advantage over those  for our students - it is free so students could easily download it at home at no cost and mess with it.  And with a little prodding, they have!  And so a new idea was born - what if we taught ALL our Master's-level courses with R?

This is just what we did this summer at Wurzweiler - all Master's-level research courses were taught using R with great success!  Over the next few blog posts, we will share a bit of what we have done with R in our classes.  We hope to inspire YOU to check out R more, too!

What do you want to know about using R in the classroom?  Let us know!

A Single-Subject Study of the Impact of Exercise on Pain and Fatigue in an Ostomy Patient

This is one last project from one of our Spring 2014 semester students. 

Our student did a really interesting project focusing on a patient with ulcerative colitis.  The patient, Beth, had had an emergency ileostomy.  Beth’s presenting problems were fatigue and pain.  It was thought that introducing an exercise regimen would address these issues. 

To measure fatigue, Beth responded to the following prompt three days each week:  “How tired do you feel today?”  This was measured with a self-anchoring scale with 0 indicating that she didn’t feel tired at all during the day and 5 indicating that she felt tired most of the time.

Beth’s pain was measured on an 11-point self-anchoring scale on the same days and times as she measured pain.  Zero indicated no pain at all and 10 indicated unbearable pain.  As Beth was primarily concerned with back pain, the prompt she responded to was a rating of her back pain, specifically.

Like our other class projects, this was an A-B design.

In a general analysis of her data, our student noted no significant trend in the baseline data for the fatigue or pain variables.  Autocorrelation was high and significant in the baseline for fatigue; however, there was not a problem with data being autocorrelated for the pain variable.

In the intervention phase, the trend in the fatigue variable was significant as was the level and magnitude of autocorrelation.  There was not a significant trend for the pain variable; however, there was a high and significant level of autocorrelation.

Therefore, to statistically compare phases, our student used the conservative dual criteria (CDC) for both of her variables.  The graph of the CDC for fatigue is shown below:

 Output from SSD for R indicated that eleven data points would be required below both the mean and regression lines in the intervention, but only two were present. 

To determine the effect size, our student used the g-index, as there was a significant trend in one of the phases.  A small effect size was observed and was in concert with Beth’s reported level of fatigue during the intervention.

To examine the pain variable, our student also used the CDC, the graph of which is shown below:

In this case, eleven data points were needed below both lines, and thirteen were obtained, illustrating a statistically significant improvement from baseline to intervention.

Effect size was measured using the d-index, and a 45.1% improvement in pain was noted since the introduction of the intervention.

This study had interesting results.  Introduction of the exercise program coincided with a decrease in back pain, but not a decrease in fatigue.  Therefore, it was suggested that Beth examine other options to relieve her fatigue.

With regard to implications for practice, our student noted that ostomy surgery often takes a toll on patients as body images are altered.  She concluded her study by saying that “more research studies are needed to break down the stigma against ostomies, to show how effective mind-body therapies can be, and to determine the types of exercise that can keep an ostomate in healthy remission.”

Nice work and interesting study!

New version of SSD for R is available

We wanted to let you know that the latest and greatest version of SSD for R is now available for download from CRAN:  Version 1.4.5

What's different about this version?  We fixed a BUG in the d-index.

This actually means, not only a correction to the Effectsize() function, but also a correction to an error in SSD for R:  An R Package for Analyzing Single-System Data.

On page 56, we display the effect size for Jenny's yelling.  With a slightly incorrect formula, we OVERESTIMATED the value for the d-index.  We initially reported a value of 2.011, but you will see below, with the correction, the actual value is 1.947.

The actual percentage change was very small, but we like precision!

And speaking of effect sizes, we will give you something to look forward - we will be adding Hedge's g to the next version of SSD for R, which will be up on CRAN at the end of July.

Finally, check out the home page of our website at:  www.ssdanalysis.com.  We are showing off the new cover of our book and provide you with links to where you can purchase it!

As always, feel free to contact us on our website:  https://www.ssdanalysis.com/Contact_Page.html

We like hearing from all of you!

Another student project - Evaluating services with a high school student

In this blog post, we will also feature the work of a student from our Spring 2014 semester.  This project was completed by Shuli Tsadok.

Shuli did her internship in a private high school, and she was working with Linda, a 10^th grade student who was new to the school.  Linda was referred for social work services because she was extremely withdrawn both with peers and with teachers; she rarely interacted with other students and did not participate in class discussions.

To measure the problem, Shuli asked Linda’s teacher to record how many times Linda raised her hand in her English class.  The baseline period was two school weeks (n=10).  During that time, Linda raised her hand, on average, 0.5 times per class period (SD=0.707).

During this assessment period, Shuli hypothesized that Linda’s withdrawn behavior was a result of social anxiety caused by extreme shyness.  To intervene with Linda, Shuli chose exposure therapy in which Linda would be put in positions to interact with others more.  Linda was assigned to a group project with other students in which she would not only have to collaborate with her classmates in a somewhat unstructured manner, but she would also have to give a presentation to the class.  Additionally, Shuli would meet with Linda in public places and encouraged her to interact with others in these settings.  Finally, Linda was given a peer tutor, which encouraged interaction with at least one other student.

After the intervention was introduced, the mean number of times Linda raised her hand in class increased to 1 (SD=0.667).

A line graph comparing phases is displayed below:

To compare the phases, Shuli noted that there was no issue of trending or autocorrelation, so she used a t-test as a statistical method.  With a calculated p-value of 0.1211, Shuli noted that there were no statistical differences between the phases.  However, with few data points and the intervention moving in the right direction, Shuli looked at the effect of the intervention.  The effect size for this intervention indicated a small degree of change.

The overall question for this course was to evaluate how well an intervention was working for a particular client.  Shuli noted that this intervention was not particularly effective for this client.  When she thought about it, she concluded that the lack of meaningful positive change could have been attributed to an unavoidable interruption in social work services combined with the fact that Linda is extremely shy.  Perhaps an extremely shy teen may have benefited from a different intervention.

Shuli noted that the “results in this study cannot be generalized because the study was only done on one client for a very brief period of time.” 

While this intervention was not particularly helpful for this client, we believe that this type of information is extremely useful to social work practitioners – it is just as important to understand what does NOT work for a client as what does!  Nice evaluation, Shuli!

What do YOU think of Shuli's project?  Feel free to contact us here and check out some of the additions to our website!

SSD for R in the Classroom: One of our final projects!

Now that the semester has ended, we thought we might show you some of our students’ final projects.  This first project was done by Catherine Faro.

The client that Catherine worked with was CW, a 21-year-old white male. CW was recently released from jail and was in treatment at a local mental health clinic with a recovery program for those dually diagnosed.  Catherine noticed that CW was not participating in his therapeutic group and sought to increase his participation.

After some research, Catherine chose to use Anger Management for Substance Abuse and Mental Health Clients: A Cognitive Behavioral Therapy Manual (Reilly & Shopshire, 2002) to intervene with this client.  It was hoped that this intervention would help CW verbalize his anger, learn adaptive coping skills, and connect better to group treatment.

To measure CW’s progress, Catherine assessed his participation at each group session.  A session was coded as 0 if CW did not participate at all, as 1 if he participated but in a non-substantive manner, and as 2 if he participated fully and constructively.

Baseline measures were taken for 11 sessions prior to beginning the intervention and for 26 sessions after the intervention was introduced.

Catherine produced the following line graph comparing the baseline to the intervention:

She also produced descriptive statistics and a box plot for each phase:         

Catherine noted that both the mean and median levels of participation rose with the intervention.

In further analyzing her data, Catherine did a trend analysis for both baseline and intervention and noted no significant trend in either phase.  Additionally, she noted no potential issues with autocorrelation in either phase.

To statistically  compare the two phases, Catherine used the binomial test to compare baseline successes with intervention phase successes in two ways.  First, “success” was defined as ANY type of group participation.  Looking at it this way, CW was successful 27.2% of the time during the baseline and 67.8% of the time during the intervention (p=0.00).  To evaluate this intervention more rigorously, Catherine then defined success as only productive group participation.  Under this more stringent definition, CW was successful during baseline 18% of the time.  During the intervention, he was successful 39.3% of the time (p=0.01).

Catherine then examined the size of the effect of the intervention using the d-index.  She found that that, with the intervention, CWs group participation improved 27.25%, a small to moderate effect.

BUT HOW DOES THIS IMPACT CATHERINE’S PRACTICE WITH CW?

Catherine concluded that the CBT techniques employed in the intervention may be having an impact on his self-reflection in group.  In her own words:

“CW became better able to discuss situations that made him angry, and examine the cognitive processes behind them and discuss his anger responses in a group setting. These two components assisted him in participating in group both in general, and in a self-reflective manner.”

Catherine said that this analysis “provides support for continued use of CBT techniques to help CW become more aware of his anger responses, and modulate them to more appropriate and constructive ways.”