I have conducted thousands of statistical consultations over the years and have worked with many novice resident researchers over that time. One cannot help but admire the spirit, energy, and motivation of young people wanting to make an impact on medicine through research. I enjoy the zeal and drive of bright people wanting to be physicians and researchers. This is a good thing!

That being said, I spend a lot of my time with novice researchers using deductive reasoning to hone down their research questions into something tangible and feasible. They come into the office with an idea that will change medicine forever and we will be cruising around the Caribbean in a year! This has never been researched before!  No one has ever done this before! Trust me, I want all of these proclamations to be true and I also want to change the face of medicine. Yet, most times it just not feasible to do so given the time, resources, participants, competencies and environment associated with the study.

I focus on a few primary areas when it comes to feasible research questions with my consultees:

1. Participant pool - Are there enough participants available in the immediate clinical or empirical environment to achieve adequate statistical power for inferential analyses? How will you recruit the participants? What are your inclusion and exclusion criteria? Inclusion and exclusion criteria may need to be modified to increase sample size.

2. Effect size - Small effect sizes require large sample sizes.    

3. Research design - Retrospective designs are always more feasible because the data already exists.

4. Communication - Research never occurs in isolation. Researchers should communicate and collaborate with their peers regarding their research projects. Attendings and academic physicians can give you ideas on how to feasibly conduct your research.

5. Time - What is the time frame for the study from inception to publication? How much time do you have to set aside for the research study? Does the completion of your research coincide with abstract deadlines of interest?

6. Power analysis - Conduct an a priori power anlaysis based on an evidence-based measure of effect to see if the study is feasible in regards to sample size needed to achieve power.

I became highly vested in the EBM literature during my second year as a professor for purposes of assessing resident/fellow/faculty/physician perceptions of EBM-based practice. I wanted to know how "knowledge gaps" were really experienced, could they be experienced, and what they did about it.  

However, I was most interested in how they accessed clinical evidence at the point of care. Some said it was readily available at bed side and others said they had their assistants run their searches for them. Others said that they read at home or had new evidence emailed to them by predetermined groups of professionals packaging evidence.

Chances are, you have many options available to you at your institution of higher learning or applied clinical practice. But most people just go straight to Google, Bing, or Yahoo to get some quick information. If this is your chosen method, then try to use Boolean operators in your quick searches to improve the specificity (finding quality evidence) of your search queries.

1. When you put quotations, " " around words or a phrase, then only those words inside the quotations will be searched. And, because there are so many nebulous words in statistics, just type the word with parentheses. Ex: "Logarithmic transformation"

When typing out a phrase or series of words in quotations, the search will follow the words in the exact order you typed them into the search engine. "how to string a guitar," or "nearest pizza place" are good examples. The search would yield specific sites and information on those two queries due to the quotations.

2. The word, OR, requires that both terms in the search query appear in the webpage or document. Using OR broadens the search yield. It can also be used to link isomorphic, similar, and interdependent concepts.

The search "statistics" OR "precision" OR "measurement" could lead to a vast number of resources linking the three constructs and can lead to new understanding of how the three interact. If you are researching an abstract construct or phenomena, the OR statement can pay vast dividends as you search the literature.

3. The word, AND, is the default of the Boolean system and is used to separate other Boolean operators. With more use of AND, the search yield will decrease. It is used to amalgamate the different "parts" of the search query together.  
The search "hotel" AND "arena" AND "paid parking," will give you a very specific search result related to close hotels with valet services that are close to the local sport arena.  

4. The words, AND NOT, will exclude anything following it in the search query. It is a good phrase to use after you have performed a few searches and have seen the same redundant sites or information pop-up. Doing this eliminates the possibility for any of the query after AND NOT from being searched.

5. Parentheses must be used with OR statements when there is another Boolean term in the search query.  

For example: "hotel" AND "arena" AND (Marriott OR Hilton)

This would find you a hotel close to the arena that was either Marriott or Hilton.  

5. The truncation, *, is a powerful search tool that will include all forms of the parent word.

For example: Isomorph* - isomorph, isomorphs, isomorphic, isomorphism, isomorphisms

These simple Boolean operators, OR, AND, AND NOT, parentheses, and truncations can yield much more specific (identifying the correct information needed from search) search finding.

I work with brilliant people in academic and clinical medicine.  These individuals are dedicated and professional people that work really hard to serve patient populations. Database management is not an everyday concern for most medical professionals. It IS an everyday concern for medical professionals that analyze data.

When working with a single research database, multiple databases, or secondary analysis of existing datasets, it is highly important to have some method for identifying individual participants with unique identifiers. Also, there must be some sort of primary "grouping" variable that distinguishes groups, datasets, or strata in more complex databases.

The unique identifiers serve the purpose of meeting the assumption of independence of observations. The "grouping" variable serves as a means for sensitivity and subgroup analyses. It further helps data analysts work with secondary, tertiary, and ancillary research questions.

The efforts you make as a researcher to have an objective and consistent method for data entry, maintenance, and analysis pays great dividends in the analysis phase of a research study. It is also the difference between it taking your statistician FIVE MINUTES to analyze your data versus FIVE MONTHS, or a happy statistical day versus a nightmare scenario.

There is a strong need in clinical medicine for adjusted odds ratios with 95% confidence intervals. Medicine, as a science, often uses categorical outcomes to research causal effects. It is important to assess clinical outcomes (measured at the dichotomous categorical level) within the context of various predictor, clinical, prognostic, demographic, and confounding variables. Logistic regression is the statistical method used to understand the associations between the aforementioned variables and dichotomous categorical outcomes.

Logistic regression yields adjusted odds ratios with 95% confidence intervals, rather than the more prevalent unadjusted odds ratios used in 2x2 tables. The odds ratios in logistic regression are "adjusted" because their associations to the dichotomous categorical outcome are "controlled for" or "adjusted" by the other variables in the model. The 95% confidence interval is used as the primary inference with adjusted odds ratios, just like with unadjusted odds ratios. If the 95% confidence interval crosses over 1.0, then there is a non-significant association with the outcome variable.  

Adjusted odds ratios are important in medicine because very few physiological or medical phenomena are bivariate in nature. Most disease states or physiological disorders are understood and detected within the context of many different factors or variables.  Therefore, to truly understand treatment effects and clinical phenomena, multivariate adjustment must occur to properly account for clinical, prognostic, demographic, and confounding variables.  

Coming from a social science background, I understand that social scientists can spend the vast majority of their time just trying to measure for the construct or behavior they are interested in. I spent a year of my life constructing a survey instrument to measure for the construct of isomorphism in clinical supervision. It is an exciting and yet daunting task to create something from nothing, and I commend social scientists that try to capture variance in human beings.

Surveys can be used to answer "unique" research questions. And by unique, things like isomorphism that exist at a very abstract or unconscious level are perceived in any number of ways to any number of people. Also remember, these types of "unique" constructs often exact a reaction of "cognitive dissonance" in your peers because they are "unknown," "different," or "weird."

All of that being said, the VERY FIRST thing you should do when conducting a survey research study is create a construct specification related to the construct you are measuring for in the proposed survey.

Remember, the survey should be written to represent just ONE construct. It is important to give an operational definition to the ONE construct. Define it in objective and measurable terms if at all possible, and use that definition as the basis for building subsequent components, content areas, and "factors." The construct specification serves as a springboard for showing how your construct exists or is theorized to exist in the context of the empirical literature. You are essentially making an argument, based on the literature in the area, that the construct can be, should be, or has not been properly assessed.

Creating a construct specification also constitutes seeking out existing survey instruments that measure something theoretically, conceptually, or empirically linked to your construct of interest. Find the "gold standard" survey instruments with the most validity evidence and seek out permission for their use in your study (if needed).  

Explicitly describe the population of interest associated with your survey. What are the inclusion and exclusion criteria for being a potential participant in your survey study? How will you go about recruiting participants? Will you use incentives?  How will you administer the survey?  Will you be able to meet sample size requirements of 150-300 for a pilot study and 300-1,000 for a validation study?

The next section of a construct specification operationalizes the content areas of your construct. Each content area should have an operational definition. Then, each component (or item) that makes up the content area should be defined and described in regards to its relevance to the construct. Lastly, give a citation from the empirical literature area to back up the argument for relevance. Do this for each component (or item) for each content area of the constuct. This can be a tedious process for more "abstract" constructs, but it is essential to provide an empirical framework/argument so that your audience can the proper frame of reference for perceiving the construct.

The last section of the construct specification is the "Table of specifications" where you given numerical designations of the percentage of the survey allocated to each content area of the construct. The number of items and content areas and their coverage within the survey must be equivalent to the make-up of the second section of the construct specification. If your construct is theorized to be composed of three content areas and one of the content areas represents 60% of the literature, then that content area should represent 60% of the items in your survey.

Going through this process is an excellent opportunity to become vested in the empirical literature and become an expert in the field. It is a time-consuming process to build a strong construct specification, but it provides a much higher quality end product.  

I designed and tested a survey instrument for purposes of my dissertation. I published the survey to an online survey administration site.

Then, I went to every website for every Counselor Education graduate program in the United States and Canada and got as many emails of students and faculty that I could find. All in all, I spent about two months of my life putting together a list of over 3.200 emails.  

After seeking out the help of the IT department, I learned how to send out mass emails to potential participants WITHOUT the emails arriving as junk or spam. Here are the steps:

1. Type every email address in ONE column of an Excel database.

2. Open up a "New Message" email and put your own email address into the To: box.

3. Click on the Cc: button to open up the menu.

4. Highlight the column of email addresses, right click your mouse, click Copy.

5. Paste the emails into the Bcc: box in the email heading.

6. Type your email out (with informed consent) and embed the link to the online version of your survey into the email.

7.  lick Send.

It's that easy!

With the dissemination of the Affordable Care Act into the healthcare system, clinical pathways and efficiency in treatment will be mainstays within medicine.  Hospitals and institutions will have to find ways to provide quality care in an economically viable fashion both for patients and the "bottom line."

Healthcare professionals are always looking at ways to cut costs.  One way that researchers will be able to assist in these endeavors is to promote the use of non-inferiority designs.  By providing a cheaper treatment that is JUST AS GOOD as the current "gold standard" treatment, costs can be reduced dramatically as it relates to clinical treatment regimens.  

There must be an empirically and clinically validated margin of non-inferiority.  Sample sizes will increase as relevant margins decrease.  These trials are most relevant when 1) secondary endpoints are better, 2) treatments are cheaper or easier to administer, 3) randomized controlled trials are not feasible or ethical, and 4) when compliance to treatment will be higher in comparison to the "gold standard."*

Better secondary endpoints, cheaper and more patient-friendly treatments, and higher rates of compliance will all be necessary as the Affordable Care Act promulgates into the healthcare marketplace.   

*Lesaffre, E.  Superiority, equivalence, and non-inferiority trials.  Bull NYU Hosp Jt Dis.  2008;66(2): 150-154.

I participate in a lot of journal clubs at my institution. As the resident "stats person," I get called by residents before they present at journal club to help them discern the statistical methods of papers. I am also asked to attend journal clubs to assist in putting statistical findings into relevant clinical contexts.

However, a pedagogical disservice is given to learners every time they are asked, "Would this evidence change your clinical practice?" This is a rhetorical question...they must NEVER let one piece of evidence change their clinical practice! This is especially true if the journal club topic that week focuses on observational or quasi-experimental designs!

The famous text by Straus et al.* stipulates that the individual trial or observational study is the LEAST FAVORABLE type of evidence to be sought out in applied clinical medicine. Systematic reviews, synopses of syntheses, and summaries are the most tangible and relevant pieces of clinical evidence when it comes to real-life patient populations.

A better use of a graduate level practitioner's time would be to seek out the highest levels of evidence in journal club, all the time. Observational studies are much more feasible for busy residents and fellows to conduct for academic requirements. But when it comes to teaching residents how to apply clinical evidence to their practice in the journal club environment, the highest levels of evidence should be used by faculty to correctly model reality-based clinical practice.    

*Straus SE, Glasziou P, Richardson WS, Haynes RB.  (2011).  Evidence-based medicine:  How to practice and teach it.  Edinburgh: Churchill LIvingston.  

We covered between-subjects and within-subjects analyses in the first Statistical Designs post. Multivariate statistics will be the focus in Statistical Designs 2.

While 90% of statistics reported in the literature fall under the guise of between-subjects and within-subjects analyses, they do not properly account for all of the variance and confounding effects that exist in reality. Multivariate statistics play an important role in empirical reasoning because they allow us to control for various demographic, confounding, clinical, or prognostic variables that mitigate, mediate, and affect the association between a predictor and outcome variable. They are also much more representative of reality and true effects that exist within human populations.

Very few if any relationships or treatment effects in physiology, psychology, education, or life in general are bivariate in nature. Relationships and treatment effects in reality ARE multivariate, diverse, and confounded by any number of characteristics. Therefore, it makes sense that researchers should be conducting multivariate statistics to truly understand human phenomena.  

With this being said, it is important to use multivariate statistics ONLY when you are asking a multivariate research question. Throwing a bunch of variables into a model without some sort of theoretical or conceptual reason for including them can yield false treatment effects and increase Type I errors. Also, these spurious variables can create "statistical noise" which detracts from a model's capability for detecting significant associations.

Choosing the correct multivariate statistic to answer your question is simple. You choose the multivariate analysis based on the outcome.

1. Categorical outcomes - Logistic regression (dichotomous), multinomial logistic regression (polychotomous), Kaplan-Meier, Cochran-Mantel-Haenszel, Cox regression (dichotomous/survival/time-to-event)

2. Ordinal outcomes - Proportional odds regression

3. Continuous outcomes - Factorial ANOVA with fixed effects, factorial ANOVA with random effects, factorial ANOVA with mixed effects, ANCOVA, multiple regression, MANOVA, MANCOVA

4. Count outcomes - Negative binomial regression (variance larger than mean) and Poisson regression (mean larger than variance)

I'm trying to get my applied and methodological research published just like everyone else out there in academia. The one thing I have learned is that all journals have different methods of submission, all with different expectations of writing styles, and varying methods for structuring the paper and citing prior research.

Submitting a research manuscript is a tedious and anxiety-ridden process. It is by no means easy, user-friendly, or logical. I have to make a judicious effort to remember the email address, ID, password, and contact information for each submission at any time. You MUST do this, because editors and reviewers will NOT do it for you.  

Upon rejection (and believe me, it is coming), you then have to completely reformat your manuscript with font and boldface changes and writing styles to meet the requisite needs of the new journal. You do this numerous times at the expense of the Impact Factor just to get published...pretty soon, what you ended with is nothing like what you started with, some months or years ago. It's no longer an original thought, but a mish-mash of editorial comments.  What is one to do?!

1. First and foremost, follow EXACTLY what the "Guideline for Authors" section tells you to do. Because top-tier journals receive many manuscripts, the author guidelines are an easy way to "weed out" manuscripts. ONE simple mistake in a citation or subheading, and they will reject it or send it back for revisions. You may have to COMPLETELY revamp the structure and writing style of the paper, but at least you have the body of the manuscript put together!

2. Also, the editors may have good ideas on how to make your paper better. If you get a rejection, integrate any pertinent changes into your manuscript. However, if you can tell that the reviewer barely read the manuscript (if at all) and gave you superfluous remarks, then you do not want to submit to that publication anyways.

3. DO NOT GIVE UP! Keep submitting your work. Do not every give up on your manuscript. If it is rejected ten times, make changes and revisions, and then send it in an eleventh time. Check the "Information for Authors" section of each publication and make sure that the journal is focused on the correct audience.  

4. Feel free to give the editors of the publication of interest an email or call. Ask if they would be interested in the study you are going to submit and if not, do they have any ideas for other potential publications? And go ahead and feel free to email them when you do not hear back from them, it is their job to get back to you.