It’s exciting to present my next interview; Karen Grace-Martin is the President and Founder of The Analysis Factor, a web-based statistical education business. Currently, Sharp Statistical Sciences is just Jesse Sharp focused on statistical consulting projects. But Karen has a team and a platform: full of live and recorded webinars and workshops as well as a forum, blog and newsletters.
The Statistically Speaking membership program offered by The Analysis Factor is a great way for solo or institutional researchers to plug into a community and some quality statistical continuing education. Here’s Karen to tell you about it! Enjoy our conversation.
I have opened my other interviews with “Do you have a ‘specialty’?” but your business provides so much more than individual consulting services. Tell me how you define The Analysis Factor and describe your main offerings. (And then go ahead and tell me, what is your specialty, if you have one?)
We approach consulting as a learning experience.
Rather than do data analysis for our clients, we work with clients to help them do their own statistics well and without the frustration of figuring it out on their own.
The Analysis Factor, as I define it, is the fact that learning statistics isn’t enough to become a skilled data analyst. Data analysis is a skill and part of becoming great at any skill is constant learning, practice, and having a mentor to guide you.
So we’ve built all of our programs—workshops, a membership community, and one-on-one consulting--all in service of helping researchers improve their skills and do great work. So everything has a training aspect, and it’s all very applied.
I don’t have a specialty topic within statistics, though I do a lot of work with linear mixed models and linear and generalized linear models. I tend to focus on the statistical topics that I see clients struggling with.
I guess you could say my specialty is breaking down complicated statistical concepts so that non-statisticians can understand and apply them.
I guess you could say my specialty is breaking down complicated statistical concepts so that non-statisticians can understand and apply them.
Marketers teach us to have a “target audience” or think about attracting the ideal client, but tell me about the person out there you would really like to help, what is she like, how can you help her?
My ideal client is a researcher who wants (or has) to do their own data analysis, wants to make sure it’s done right, and wants to really understand what to do and why. They recognize the benefits of improving their own skills and in the importance of doing their data analysis well.
It doesn’t matter how much training she has already had, as long as she is willing to learn.
Our one-on-one consulting services are all done through web meetings, where we can discuss how to approach a data analysis, how to set up data, or how to interpret results. Whatever gives her the boost to keep moving forward.
When someone needs to learn a specific analysis deeply, we offer 8-12 workshops a year on different statistical topics. These are very applied and include a ton of support materials.1 They’re all run live via webinar to keep the costs and time commitment down for participants.
But my favorite and most unique service is our Statistically Speaking membership community. It gives researchers a place to get accurate, trustworthy statistical advice as well as ongoing statistical training. Members are supported by a team of statistical consultants who run weekly online Q&As, teach monthly training webinars, and answer questions on a private forum. Because we set it all up online we’re able to keep the costs much, much lower than one-on-one consulting.
We have nearly 500 members in the program and it’s been running strong for 5 years.
Unfortunately, many people don’t get good statistics or research methods training before they find themselves working on an investigation that requires complex techniques—which is why we need The Analysis Factor—describe how you help these clients make the connection between the math and their area of application.
I’ve been helping data analysts now for almost twenty years and what I’ve found is that even good statistical training isn’t enough. Data analysis is just too complex of a task and statistics is too wide of a field for anyone to know everything. And of course, knowing the theory is only part of applying it.
I do a lot of teaching through example and breaking the statistical topics down into what makes logical sense to people who haven’t had a lot of math. I explain things in words rather than equations. And when equations are unavoidable, we walk through them piece by piece. A lot of it is meeting clients where they are. And of course, the best example to teach with is the analysis someone is actually doing.
I never just give someone an answer on what to do without making sure they understand why.
You are still very hands on and conduct some of the live workshops. What’s the best part of teaching for you?
I really am a teacher at heart. My absolute favorite part is when someone has been struggling with a concept or analysis and they finally get it. You can hear it in their voice—that change from frustration to clarity as the light bulb turns on. That really makes my day.
I recently had a client in linguistics who had to do some really complicated models. He recently told me that the level of clarity he gained in working with me really helped him in his job talks and he got his dream job at the university of his choice. He said the audience was so impressed he could explain the models so that even the humanities faculty could understand it. I was so proud of what he accomplished.
You are currently revamping your website, and it looks great, plus you have a new Linear Models workshop, what else is new for you in 2018?
Last year we grew a lot. We brought on two new staff members and three new instructors. So yes, we have the new Linear Models workshop that just started last week. We are also planning two other new workshops for the fall—on Survival Analysis and on Generalized Linear Mixed Models. Both are topics we’ve had a lot of requests for.
We’re also working on slowly improving many of the things we’ve been doing all along. The website is just one example—we’ve been working on the looks of the site, and we’re also working on making sure new articles are posted more often, we’re running more free webinars, and that relevant information is easier to find. Our new team members have made all the difference in what we’re able to do.
We really have a lot of fun and it’s so rewarding to know you’ve really helped someone with a real problem they’re struggling with.
Wow, thanks Karen! To find out more visit The Analysis Factor home page.
Check out the Repeated Measures workshop which begins again on February 27, as well as other live workshops already scheduled for 2018.
If you are unsure whether a workshop is right for you read the entire page description and if you still have questions, Karen or her team will answer them.
Full disclosure: I am a paying member of Statistically Speaking and have attended the live repeated measures workshop and several recorded webinars.
More about coopetition: https://en.wikipedia.org/wiki/Coopetition
Previous post in this series: Interview with Omega Statistics
Much of my time as an applied statistician and analyst is working on other people’s problems using already existing data, so there is very little opportunity to guide the formulation of questions and the subsequent research methods and design. However, data gathered without the proper forethought is frequently not as useful for answering the chosen question.
The challenge is not in finding problems that need solving. A large portion of my working adult clients began their research by focusing on a problem they observed in their own professions.
So, what is the problem with problem statements?
In this context, problem statements are gap statements and gap refers to something missing in the research or body of knowledge. For example, in looking at workplace discrimination against those suffering from traumatic brain injury (TBI) there is a practical issue of people who want to work being un- or under-employed but the gap in the research addressed was the missing perspective of the TBI individuals themselves. The problem is the bigger context, employment discrimination against TBI survivors, the gap is the focus of the research, getting the perspective of TBI individuals.
Here is an example of a well written problem statement using a short piece presenting a summary of research in elder abuse (pdf).
Elder abuse represents a serious and growing public safety and public health concern. Although elder abuse has been shown to have negative eﬀects, very little is known about the possible consequences of elder abuse over time, particularly long-term outcomes. The National Institute of Justice (NIJ) has been critically committed to expanding knowledge regarding this issue. Through NIJ’s involvement, we have not only been able to investigate the link between elder abuse and negative outcomes, but also on tangible solutions for victims, specifically what aspects of elder adults’ lives can be changed in prevention and intervention eﬀorts. This article describes a series of studies that have pointed to the critical role that social support may play in preventing and ameliorating the negative eﬀects of elder abuse.
The first sentence above describes the problem, the second mentions the research and shows a gap in knowledge. The last sentence shows that the research we are about to read has been done to fill the gap.
A complete problem statement:
If you are working on your dissertation then your body of knowledge is your complete area or domain of study, including the associated theories, concepts and terminology. In the domain of project management, for example, the long-standing framework is literally the Project Management Body of Knowledge or PMBOK (copyright PMI).
Here is a wonderful page with many good examples of gap statements, including practice examples for you to do:
I recommend spending as much time as needed to create the best problem/gap statement you can. From this will flow the research questions you create and the methods and designs you employ. In other words, you want to define a problem that leads to a question that you can actually research and test.
In this chapter we consider the essential first stage of survey research—devising and developing a precise research idea which is capable of leading to valid, reliable, interpretable and useful results. - Survey Research, Chapter 2
In his book Survey Research, Roger Sapsford uses chapter 2 to cover “What’s the problem? Developing ideas.” Sapsford opens the chapter with a discussion of the need to think about the final product of the research at the outset. “So, thinking about what we shall have to put into the final report has given us a list of questions we need to ask at the very beginning:
Here is a possible visualization. Your motivation is a problem within your domain. You begin to research the problem and discover that there is an aspect not covered in an academic or systematic fashion—the gap. From the problem and gap, you devise the problem statement. From that flows your research question or questions and from that flows your dissertation. At the same time you want a problem statement that is going to grow from a few paragraphs into a full dissertation!
Finally, get yourself this 53-slide presentation from Marilyn K. Simon over at Dissertation Recipes on Creating the Problem Statement, The Key to Your Dissertation or Research Project. She has practical, specific advice with important reminders such as the need to watch for statements that leave the reader going “So what?”
The next post in the quality dissertation series will cover how we get from the problem statement to the actual research questions. Ask a question or share your tips about problems and gaps in the comment section below or contact me.
Sapsford, Roger. Survey research. Sage, 2006.
If you haven't heard of or thought of using mind-mapping check out these links. This technique is useful in many situations:
When I launched Sharp Stat Sci’s website an intended feature was Chart of the Month, linking to or creating an interesting visual and discussing what makes it an effective graph. But, I didn’t keep up with that!
Instead I begin this series on “pretty pictures.” In future posts I will discuss aspects of charts and graphs that make them necessary for almost any analysis. For now, here are is a selection of outlets with compelling regular visual presentations and two graphical roundups from 2017.
Leave a comment on where you find good and bad graphs and what you would like to know more about when it comes to visualizing data.
The chart above was produced by the New York Times along with the American Statistical Association as the sixth of the monthly “What’s going on in this graph?” feature. Several colleagues have been tweeting and highlighting this series since it began. Intended for those formally labelled “teachers” and “students” it is of interest to anyone who wants to picture data. The next live discussion takes place Feb. 13, 2018.
The Economist (limited access without subscription) has a section devoted to “charts, maps, infographics and interactive data features” and a daily chart. This recent chart shows that female economists not only write more clearly than their male counterparts, but they must work harder to continue being published. I like this particular graph because all of the elements are clear, what each axis and observation represents can be derived from the graph without the need for further text.
The Kaiser Family Foundation presents a chart of the week in healthcare, this January 2018 chart is a response to allowing work requirements for Medicaid. This shows that even without this provision, most Medicaid recipients already work.
More charts to check out:
This is a pretty non-political round-up of charts on various topics. Of interest to me is that most of these charts are very simple line charts, but for the most part effective at conveying the topic. Three, four and eight show change over time by making all the starting values one or 100%.
A more global view of 2017 with a wide variety of chart types.
44 Kinds of charts: http://blog.visme.co/types-of-graphs/
Pure joy of visualization, a modern use of data:
Read the very interesting article:
Project website: https://www.oddityviz.com/
Who knew coopetition was already a thing? Cooperative competition is a great way to do business!
The prior post highlighted two free webinars on the dissertation process. This post highlights Elaine Eisenbeisz of Omega Statistics, the statistician who brought you those great resources. Elaine is a colleague and cooperative competitor. She’s been at this awhile and I appreciate her encouragement and support. For those I can’t help, Elaine is someone I recommend without reservation.
In October 2017, Voyage LA did a full feature interview you can read here. I want to focus more specifically on how Elaine can help you with your dissertation.
Do you have a “specialty,” either a research method or class of techniques or subject area or…?
John Tukey said a statistician gets to play in everyone’s backyard, so I help researchers in many fields. But I specialize in two niches per se: clinical/biotech research and dissertation research. I work with pharmaceutical companies, biotech companies, and people who are working on their PhD dissertation research. That might sound diverse. But I see it as “helping researchers move the knowledge base forward.”
Following on that, marketers teach us to have a “target audience” or think about attracting the ideal client, but tell me about the person out there you would really like to help, what is she like, how can you help her?
My ideal client is someone who cares about their research being done well and sees the value in what I offer them. More specifically, a researcher who knows the importance of a well-designed study and a well-executed data analysis.
I care deeply about setting up a study in a way that enhances success. I can’t guarantee success. But I can help researchers to know if their study is feasible and to see options and possible pitfalls upfront. I often get data sets and requests for help after the data have been collected, and although I can help patch up a poorly designed research, I’d rather come on board earlier in the design stage to make sure the information is gathered properly and adequately.
My target audience are researchers who are struggling with their study design or data analysis. Perhaps a dissertation chair has suggested the researcher hires a statistician or dissertation coach. Or a biotech start-up is in need of study design or data support and isn’t quite at the point of hiring a statistician full-time. Sometimes, I am just a second set of eyes on a project. But I like projects where I get to be more hands-on in the process.
When you are working with a client what is your favorite part of the dissertation or dissertation process?
When I get to send the congratulation card addressed to "Dr. Client, PhD"!
I also love it when the client learns how statistics are related to their specific study. I tell my dissertation clients that they don’t need to know the whole of statistics. Just the stats relevant to their study and findings. To me, statistics is most useful and powerful when used in application.
I care deeply about setting up a study in a way that enhances success.
My colleague at Omega Statistics is doing a series of free webinars providing useful, in-depth tips for a successful dissertation process. Sign up for the November 8, 2017 webinar to learn more about the difference between the concepts of method and design and how your choices affect the entire project. (Update: here is the recording.)
Also, check out the recording of the first webinar, Charting Your Course: Navigating the Literature Review which clearly describes what a lit. review is (and isn’t) and supplies essential ideas for search, style and organization. And sign up for the Omega Statistics newsletter while you are there!
Finally, head over to the Dissertation Angels site and check out the post on Choosing a Dissertation Consultant, Coach or Editor.
This is the first in a series dedicated to highlighting high-quality dissertation advice and consultants amid the increasing rise of low-cost, low-quality vendors. If you are looking for specific advice or assistance or if there is a topic you would like to see covered, leave a comment or contact me.
As we head into the end of summer and gear up for fall, here are some updates on topics we have previously covered.
American Opioid Epidemic:
September and October of 2016, we did a four part series on the opioid epidemic, a topic I have been personally and professionally following since 2011. In the time since those posts, the problem has prompted outlets of the news on line and on air to create special series dedicated to covering addiction and it's impact. KCRA 3 is the local NBC news affiliate that covered the over 50 deaths that occurred here in April 2016 due to imitation Norco laced with fentanyl.
The epidemic has the dubious distinction of its own Wikipedia entry.
While the Whitehouse did issue a press release to declare the situation an emergency, as the President’s Commission on Combating Drug Addiction and the Opioid Crisis requested, as of this posting nothing has come of it. Vox has a discussion over what might be doable.
Where do we go from here? STAT News lays out a grim forecast.
If you have a research project in this area and need a statistician, please, get in touch for a quote.
We first talked about this in September 2015 and followed up in July 2016 and the topic continues to get more airplay. The scientific society Sigma Xi has started a series of conversations on scientific reproducibility, seeking ideas from membership on how we might approach the problem. The post has already generated at least one impassioned reply that sets the issue in an even broader context.
The ASA has also released a statement aimed at funding agencies for ensuring reproducible research.
And, on a more practical level, Joris Mueller posts on Ten Simple Rules for Reproducible Research, and Tim Grossenbacher presents A Truly Reproducible R Workflow.
Significance and p-values:
On March 9, 2016, we posted about the American Statistical Association’s Statement on p-values. I still get a geeky-thrill reflecting on this milestone! As it happens, ASA has organized the Symposium on Statistical Inference around “Scientific Method for the 21st Century: A World Beyond p < 0.05.” The goal is to go beyond learning about existing sound statistical practice but to advance positive change in “communicating and understanding uncertainty, and decision-making.” Find out more here.
At a more practical level, here is short and simple guide for those who may find themselves doing research without having had an introductory college course. Using a simple, practical example it walks through the basics of odds ratios, confidence intervals and p-values.
Finally, a recently posted piece on Vox by Brian Resnick gives a lengthy overview of what p-values are and covers the current discussion around reproducibility and statistical significance, definitely worth a read.
On May 10, Sharp Stat Sci celebrated two years in business. For some days now, I have been receiving warm notes of "Congratulations!" from my network and I wanted to take a moment to acknowledge how much the encouragement means to me. One of the biggest shifts in going from being a member of an analytic team in a large organization to being a sole proprietor is the sudden loss of the built-in day-to-day contact with other folks.
The virtual network that remained and the new one I have built as I have gone along has resulted in meaningful relationships and many learning and consulting opportunities. First, however, public thanks to my professional coach, GR, is long overdue. She shared all of her accumulated practical knowledge in starting a business with me—from getting a website to getting an ETIN to tips on creating and monetizing content, marketing and setting rates—and was a cheerleader in those first very slow months.
Along the way, I was fortunate to stumble into a small group of independent consultants who decided to meet weekly to talk about our business issues, goals and share ideas and resources. Much of my success in the last 12 months is due to my participation in this group, and some referrals! I can’t thank this group enough.
My clients have been amazing and diverse; almost all are remote, of course. For many of them these are important projects they have invested a lot of time in and helping them successfully complete is really an honor. I’ve been able to help dissertation clients most recently in business administration, education, public health and economics. Additionally I have recently been able to help some university researchers and an industrial trade organization.
So, thank you, everyone, for your support. And if you have some analytic problems to solve let me know.
At the time of writing there are still 56 2/3 % of days in April left to celebrate! One of the many reasons I like to promote the practice of statistics is the real opportunities for meaningful careers.
Since graduating in 1997 with an M.S. in Statistics with an emphasis in Biostatistics, I have never lacked for job choices. Just starting or mid-career it is never too late to add analytics to your repertoire!
Check out the Math and Stat awareness homepage.
The American Statistical Association offers up this article: Celebrate the Significance of Mathematics and Statistics at Amstat News
My statistics alma mater, UC Davis posts 7 cool facts about math and statistics. Here is a sample:
Statistics Majors are on the Rise
Finally, here are a few statistics classes you can look into. Summer Statistics Courses 2017.
(list credit Janet R, Clinical Assistant Professor)
So get out there and do some math!
A number of things have kept me from posting lately, including plenty of involved and interesting projects. Additionally, it has been tempting to continue to write about the ongoing opioid epidemic but one of my goals is to bring information perhaps more relevant to my target audience. The best business advisors tell entrepreneurs to carefully define their audience and potential clientele, and with good reason. You can’t be all things to all people and maintain high quality standards as well.
When I did my opioid series a friend and colleague asked me whom I was writing for and all I could tell her, in that moment, was that I was writing it because it is interesting to me and I am interested in people who find it interesting. (And there is no doubt about the gravity of the problem.) But her point was important and she gave me a lot to think about when it comes to how I want to use this platform.
So who is my target audience? I love to work with people who need help with their statistics and data analysis in order to get something done. Sample size and power or methodology to get a grant accepted. Statistical analysis to get that report completed, finally! Or finally getting through the data analysis for your dissertation or thesis—with the conviction you will successfully defend and finish. I also offer to write chapters 3, 4 and 5 with you so you do not have to go it alone. Contact me for more information at email@example.com.
Even as I work on my focus, I cannot give up my broad interests and wanting to share them. Here are a few links to some of the many interesting conferences in data science, statistics and healthcare on tap for this year.
Data Camp’s 2017 Conference Guide
Including these two:
#ODSC – Open Data Science Conference, multiple dates and locations
Insurance Nexus USA 2017 (Chicago, IL: March 14-15, 2017)
This is the fourth and final post in a series on opioids, arguably the largest public health crisis of the early 21st century. The goal is to give context through education and facts to supplement the daily news about overwhelmed neighborhoods, police and emergency rooms battling overdoses. Part 1 can be found here, part 2 here and part 3 here.
In the first part of this series we introduced the opioid epidemic from the viewpoint of Workers’ Compensation research into the problem which began in earnest at least five years ago. This research was primarily focused on physician prescribing patterns and the growth both in the quantity of prescriptions and the amounts being prescribed, including increases in more potent narcotics such as fentanyl.
The second part defined narcotics and the concept of morphine equivalent doses or MEDs (also known as morphine equivalent amounts, MEAs, and morphine milligram equivalents, MMEs). Additionally, we looked at the general pharmacology of opioids and finally we discussed three factors contributing to the narcotic problem:
Part three was an extremely brief look at stories of overdoses and deaths easily found in the daily news and on medically oriented websites. In addition, we pointed out a just published study on the opioid epidemic through the lens of private health insurance.
This final post looks at the actions that are being taken and that can be taken to stem the tide of addiction and abuse and restore well-being to individuals and their communities. Not surprisingly, much of this centers on providers and their patients. But we should keep in mind we each have a role—these providers and patients are our friends, family and colleagues.
In late August the Surgeon General took the unprecedented step of sending a letter to all providers—about 2.3 million physicians—addressing the opioid crisis. The letter (and a good deal of other information) can be found at Turn the Tide Rx. Here is the second paragraph of the letter, commenting on the issues we discussed previously:
It is important to recognize that we arrived at this place on a path paved with good intentions.
Additionally, the letter mentions a CDC report that indicates no real increase in pain-related diagnoses that might account for some of the epidemic. Finally, the letter was sent including a pocket-card summary of the CDC Opioid prescribing guidelines.
The card can also be found on the Turn the Tide website along with two other useful summary cards—much easier to read than the technical prescribing guidelines. The first card discusses non-opioid pain treatments for chronic pain. One thing we have not addressed very well in this series is the kinds of pain for which opioids are and are not appropriate. Opioids have a place in the treatment of moderate to severe acute episodes of pain as well as in cancer treatment, palliative care and end-of-life care. The evidence is mixed for the long term use of opioids to treat chronic pain (chronic is defined as greater than 3-months).
Non-narcotic options include non-drug options: changes in diet, mindfulness, massage, physical therapy, activity and cognitive behavioral therapy, among others. Non-steroidal anti-inflammatory drugs (NSAIDS) such as aspirin, ibuprofen and naproxen can be obtained over-the-counter and some can be taken at “prescription” level doses—with a doctor’s advice of course. Acetaminophen is not an NSAID and should not be taken at doses higher than recommended as it is proven to cause liver damage. Many prescription narcotics are combination products that contain acetaminophen which should be taken into account. The FDA has a small and helpful piece oriented towards patients and consumers that covers some of this information.
Other non-narcotic possibilities discussed are tricyclic antidepressants (e.g. amitriptyline), serotonin and norepinephrine reuptake inhibitors (SNRIs, e.g. duloxetine) and anticonvulsants (e.g. gabapentin). Note that many of these should not be taken along with opioids due to serious side-effects.
Helios, a workers’ comp subsidiary of Optum recently published the second-edition of their Clinical Info newsletter which discusses some of the information supplied above. The newsletter begins with a state-by-state discussion of actions being taken in the WC industry to curb the misuse and abuse of narcotics. Many states are mandating the use of PDMPs (prescription drug monitoring programs). This is true for both WC and for group health.
These are databases that doctors are required to check for existing prescriptions their patients may have received from another doctor. Here in CA Gov. Brown just signed a bill mandating that the existing database be used when a doctor is going to prescribe a controlled substance. Health Affairs recently published a study showing an association between a reduction of opioid-related deaths and PDMPs.
Other states are putting regulations in place that limit the quantities of an opioid that can initially be prescribed for acute pain, say after surgery, or for chronic pain. Another option being pushed, both by the Surgeon General and in WC is lower doses since it has been shown that abuse and addiction risks increase at higher doses while there is some evidence that the associated pain relief does not.
Not only do prescribers need to learn best practices but individuals need to begin to understand that prescription drugs for pain are not curative but simply symptom relievers and that there are other—non-addictive—possibilities. There is no doubt that severe or chronic pain can diminish quality of life. However, there is some evidence that opioids can make pain worse. It is our responsibility to question our doctors and to demand that they too learn about all of the pain-treatment options available.
Some researchers are working on non-addictive pharmaceuticals for pain relief. While this is hopeful it still ignores the reality of pain and the possibility of non-pharmaceutical solutions. The website, PainAction.com, while partially sponsored by at least one drug company, is oriented towards people who suffer from various forms of chronic pain and supplies helpful educational materials as well as links to other resources.
Jesse Sharp is an expert in the analysis of health care data. Passionate about data and the ethics of analysis he writes on topics related to medicine, public health and statistics. More...