Page 1 of 9

Advances in Social Sciences Research Journal – Vol. 11, No. 8

Publication Date: August 25, 2024

DOI:10.14738/assrj.118.16706.

Troy, J., Grambow, S., Neely, M., Pomann, G.-M., Davenport, C., Ashner, M., & Samsa, G. (2024). Rationalizing the Mathematics

Requirements for a Master of Biostatistics Program: A Case Study and Commentary. Advances in Social Sciences Research Journal,

11(8). 265-273.

Services for Science and Education – United Kingdom

Rationalizing the Mathematics Requirements for a Master of

Biostatistics Program: A Case Study and Commentary

Jesse Troy

Department of Biostatistics and Bioinformatics,

Duke University, Durham NC, USA

Steven Grambow

Department of Biostatistics and Bioinformatics,

Duke University, Durham NC, USA

Megan Neely

Department of Biostatistics and Bioinformatics,

Duke University, Durham NC, USA

Gina-Maria Pomann

Department of Biostatistics and Bioinformatics,

Duke University, Durham NC, USA

Clemontina Davenport

Department of Biostatistics and Bioinformatics,

Duke University, Durham NC, USA

Marissa Ashner

Department of Biostatistics and Bioinformatics,

Duke University, Durham NC, USA

Greg Samsa

Department of Biostatistics and Bioinformatics,

Duke University, Durham NC, USA

ABSTRACT

Within a 2-year Master of Biostatistics program, we reconsidered the mathematics

requirements for admission, and also the premise that the most distinguishing

feature of a top candidate is deep exposure to mathematics. Our assessment took

place within a broad curriculum review intended to enhance alignment: aligning

programmatic goals with job skills, aligning the use of mathematics within our

curriculum with programmatic goals, aligning our admission criteria with our

curriculum, and aligning our application materials with these admission criteria.

We developed a specific list of mathematical skills required by the curriculum, and

are revising the application materials to include self-report on applicant's exposure

to and functional mastery of those skills. We illustrate how functional mastery is

operationally defined. Our criteria for identifying top candidates was broadened to

include those with especial skills in analytics, biology and/or communication (i.e.,

Page 2 of 9

266

Advances in Social Sciences Research Journal (ASSRJ) Vol. 11, Issue 8, August-2024

Services for Science and Education – United Kingdom

the three conceptual pillars of our program). Deep mathematical training is one of

many ways to become a top candidate. Within STEM fields such as biostatistics, and

despite the commonly held assumption that admission criteria should emphasize

depth of mathematical training, a systematic analysis suggests that this assumption

imposes a gratuitous requirement on applicants. Reconsidering this assumption

can help remove unnecessary barriers, reduce challenges, and support success for

aspiring or emerging biostatisticians from diverse or multifaceted backgrounds. It

is one way to contribute to a more inclusive and equitable learning environment in

our STEM discipline. We believe similar programs might benefit from performing

this type of analysis and reflection.

Keywords: Biostatistics, curriculum review, mathematics requirements, STEM pipeline.

INTRODUCTION

Lorem A basic premise of program development is that of constructive alignment [1] -- for

example, curriculum content should be aligned with overall educational goals, and admission

criteria should be aligned with curriculum content. Moreover, educational goals should be

realistic. For example, the curriculum of a professional master's program should be focused on

tasks that graduates will accomplish on the job (currently, and in the future), and gratuitous

requirements should be avoided.

Here, we consider a 2-year Master of Biostatistics (MB) program. Originally approved as a

terminal professional degree program, it now serves two types of students. Approximately 30%

immediately proceed to doctoral study after graduation. The other 70% enter the workforce. A

distinctive feature of the MB program is its broad focus, consistent with a mission to train

students who will contribute to interdisciplinary team science, encompassing not only

Analytical skills but Biological knowledge and Communication skills as well (i.e., the "ABCs of

biostatistics") [2-10]. Graduates often take positions with significant responsibility and

autonomy, and our intention is to train students for this type of "high-end" application.

The current case study and commentary pertain to the alignment between curriculum content

and admission criteria. Because a biostatistics curriculum is mathematically intensive students

must be "good at math", and our question is how that construct can be best operationalized.

Previously, our measure of this construct was based on grades and course lists. At a minimum,

we required calculus through multi-variable calculus and linear algebra (i.e., the two main

languages of instruction), with the default expectation that the multi-variable calculus grade

was an A. Moreover, top candidates were expected to have additional coursework in real

analysis, differential equations, etc. In other words: (1) the primary criterion for basic

competency was demonstrated excellence in multi-variable calculus; and (2) a requirement for

being considered as a top candidate was deep training in mathematics, and the deeper the

better. Other master's programs apply similar admission criteria.

We continue to believe that mastery of multi-variable calculus is an excellent measure of basic

competency. However, the seemingly intuitive criteria for identifying top candidates had some

unintended consequences. We begin by describing those consequences, and then how our

admission criteria are being modified in response. We believe similar programs could benefit

from performing this type of analysis and reflection.

Page 3 of 9

267

Troy, J., Grambow, S., Neely, M., Pomann, G.-M., Davenport, C., Ashner, M., & Samsa, G. (2024). Rationalizing the Mathematics Requirements for

a Master of Biostatistics Program: A Case Study and Commentary. Advances in Social Sciences Research Journal, 11(8). 265-273.

URL: http://dx.doi.org/10.14738/assrj.118.16706

UNINTENDED CONSEQUENCES

Our response to these unintended consequences began with a comprehensive and ongoing

curriculum review. An initial step was to identify the skill set required by our graduates and to

revise the curriculum so that (1) each important skill was addressed; and (2) less important

skills were deemphasized, eliminated, or made optional. This step led to big-picture changes

such as adding and dropping courses, revising the mapping of courses into tracks, etc. These

changes supported better alignment between the overall curriculum and overall educational

goals (i.e., made concrete by the desired student skill set).

Next, we reviewed how effectively important skills were taught within specific classes (e.g.,

active learning approaches were encouraged). This was the point at which we assessed the

requisite degree of mathematical coverage. For example, our first inference course was revised

to replace some (but certainly not all) calculus with demonstrations via simulation, and to

replace formal proofs with other intuition-building exercises [9]. This supported the second

element of alignment: namely, better alignment between course goals and course content

(especially, mathematical content).

Once this initial process of course revision was completed, we considered how mathematics is

used within individual courses. Upon reflection, two differing skill sets are needed: (1) the

general ability to use mathematics to formulate and solve statistically-based problems; and (2)

mastery of specific prerequisite knowledge required by our program. The second and more

specific skill set was simpler to operationally define.

Regarding prerequisite knowledge, we reviewed the courses that a job-bound student would

be likely to take (thus, for example, selecting elective courses with an applied focus) and

recorded the specific applications of mathematics. To our surprise, this list (Appendix 1) is not

voluminous. It contains deep working knowledge of some basic concepts, such as the nature of

functions. It also includes a relatively small subset of the material typically covered in an

undergraduate course in calculus which focuses on the trajectory of objects as they move

through space and similar concepts from physics. Additionally, it includes (1) facility with

differentiation and integration, which are required (among others) for manipulating the

probability distributions used in statistics; and (2) ability to manipulate functions -- especially,

to maximize and minimize them.

A similar phenomenon was observed for linear algebra -- rather than the entire content of a

typical linear algebra course, the information that is most important to our program is limited

to (1) the ability to manipulate matrices (e.g., to add them, to invert them); (2) the ability to

state statistical models in matrix terms (i.e., as this allows ideas to be presented succinctly and

in greater generality); and (3) the ability to project high-dimensional spaces onto lower

dimensions (i.e., this being critical to statistical modeling, and requiring facility with content

such as eigenvalues and eigenvectors).

Regarding the first skill set, the underlying construct (i.e., general ability to use mathematics to

formulate and solve statistically-based problems) was straightforward to state, but not

necessarily trivial to assess. Presumably, having good grades in undergraduate courses in

mathematics, or which extensively use mathematics, is a necessary but not fully sufficient

condition for assessing this construct, because students can master the technical elements of

Page 4 of 9

268

Advances in Social Sciences Research Journal (ASSRJ) Vol. 11, Issue 8, August-2024

Services for Science and Education – United Kingdom

mathematics as an exercise in symbol manipulation but not truly understand the underlying

concepts (and, thus, not be able to extend their knowledge to different problems). In any event,

we had created a list of what students should be able to do: namely, use calculus to accomplish

a specific set of tasks, use linear algebra to accomplish another specific set of tasks, and apply

mathematics more generally to define and solve statistical problems. For each of these

constructs, our admission materials could only provide a rough surrogate, albeit a surrogate

that we hoped to improve.

RESPONSE

Our response to these unintended consequences began with a comprehensive and ongoing

curriculum review. An initial step was to identify the skill set required by our graduates and to

revise the curriculum so that (1) each important skill was addressed; and (2) less important

skills were deemphasized, eliminated, or made optional. This step led to big-picture changes

such as adding and dropping courses, revising the mapping of courses into tracks, etc. These

changes supported better alignment between the overall curriculum and overall educational

goals (i.e., made concrete by the desired student skill set).

Next, we reviewed how effectively important skills were taught within specific classes (e.g.,

active learning approaches were encouraged). This was the point at which we assessed the

requisite degree of mathematical coverage. For example, our first inference course was revised

to replace some (but certainly not all) calculus with demonstrations via simulation, and to

replace formal proofs with other intuition-building exercises [9]. This supported the second

element of alignment: namely, better alignment between course goals and course content

(especially, mathematical content).

Once this initial process of course revision was completed, we considered how mathematics is

used within individual courses. Upon reflection, two differing skill sets are needed: (1) the

general ability to use mathematics to formulate and solve statistically-based problems; and (2)

mastery of specific prerequisite knowledge required by our program. The second and more

specific skill set was simpler to operationally define.

Regarding prerequisite knowledge, we reviewed the courses that a job-bound student would

be likely to take (thus, for example, selecting elective courses with an applied focus) and

recorded the specific applications of mathematics. To our surprise, this list (Appendix 1) is not

voluminous. It contains deep working knowledge of some basic concepts, such as the nature of

functions. It also includes a relatively small subset of the material typically covered in an

undergraduate course in calculus which focuses on the trajectory of objects as they move

through space and similar concepts from physics. Additionally, it includes (1) facility with

differentiation and integration, which are required (among others) for manipulating the

probability distributions used in statistics; and (2) ability to manipulate functions -- especially,

to maximize and minimize them.

A similar phenomenon was observed for linear algebra -- rather than the entire content of a

typical linear algebra course, the information that is most important to our program is limited

to (1) the ability to manipulate matrices (e.g., to add them, to invert them); (2) the ability to

state statistical models in matrix terms (i.e., as this allows ideas to be presented succinctly and

in greater generality); and (3) the ability to project high-dimensional spaces onto lower

Page 5 of 9

269

Troy, J., Grambow, S., Neely, M., Pomann, G.-M., Davenport, C., Ashner, M., & Samsa, G. (2024). Rationalizing the Mathematics Requirements for

a Master of Biostatistics Program: A Case Study and Commentary. Advances in Social Sciences Research Journal, 11(8). 265-273.

URL: http://dx.doi.org/10.14738/assrj.118.16706

dimensions (i.e., this being critical to statistical modeling, and requiring facility with content

such as eigenvalues and eigenvectors).

Regarding the first skill set, the underlying construct (i.e., general ability to use mathematics to

formulate and solve statistically-based problems) was straightforward to state, but not

necessarily trivial to assess. Presumably, having good grades in undergraduate courses in

mathematics, or which extensively use mathematics, is a necessary but not fully sufficient

condition for assessing this construct, because students can master the technical elements of

mathematics as an exercise in symbol manipulation but not truly understand the underlying

concepts (and, thus, not be able to extend their knowledge to different problems). In any event,

we had created a list of what students should be able to do: namely, use calculus to accomplish

a specific set of tasks, use linear algebra to accomplish another specific set of tasks, and apply

mathematics more generally to define and solve statistical problems. For each of these

constructs, our admission materials could only provide a rough surrogate, albeit a surrogate

that we hoped to improve.

MODIFYING THE ADMISSION PROCESS

Considering the above, we have made two major changes to the admission process. The first is

an attempt to better translate course titles into an assessment of the mastery of the specific

items in Appendix 1. Toward this end, we had previously asked that applicants describe the

level of rigor associated with their mathematically-related courses, and many applicants had

additionally copied course descriptions from a catalog. Starting with the next application cycle,

we will enhance this report by (1) asking students to self-rate their facility with the items in

Appendix 1; and (2) provide example responses that illustrate how we conceptualize mastery

(see Appendix 2). In other words, we will ask applicants to provide more specific self-reports

about their mastery of the content that will be important to their success in our program, and

also provide examples which help them calibrate these reports.

The second change was to reconsider what it means to be a top candidate. Rather than simply

ranking students on the number of mathematically-related courses taken as an undergraduate

(1) we continue to require all successful applicants to demonstrate basic competency in multi- variable calculus and linear algebra; and (2) for applicants satisfying this first condition, we

now define a top candidate as someone with especial skills in analytics, biology and/or

communication (i.e., the three conceptual pillars of our program). In other words, additional

mathematical training remains one way to become a top candidate, but isn't the only one.

Indeed, we had previously been making such judgments on an ad hoc basis, and have found that

doing so in a systematic, transparent and reproducible basis helps to simplify the application

review process as well as making it more consistent.

DISCUSSION

We have described the process by which our admission requirements around mathematics

were systematically reviewed and updated. Our intention is to more effectively align the use of

mathematics in our curriculum with broader programmatic goals, to better align our admission

criteria with how mathematics is used, and to better align the information in our application

materials with these admission criteria. Reconsideration of admission criteria was a final step

in a broader process of curriculum review. We were encouraged to find that the specific list of

mathematics requirements was of modest length, and favored depth of knowledge over

Page 6 of 9

270

Advances in Social Sciences Research Journal (ASSRJ) Vol. 11, Issue 8, August-2024

Services for Science and Education – United Kingdom

breadth. We also recognized a mismatch between this list of mathematics requirements and the

information currently available on our application (e.g., lists of undergraduate math-related

courses and their grades), and are revising our application materials accordingly. In part

because of the ability to perform web searches, to use generative artificial intelligence, etc., we

do not directly evaluate mathematical skills within the application (e.g., we don't ask applicants

to answer questions such as those in Appendix 2), but instead provide information about our

interpretation of functional mastery of mathematical skills to help support their own self- evaluation.

An additional benefit of this review is that potential applicants will receive more specific

information about the preparation they need to succeed in our program -- for example, to assist

them in their selection of undergraduate courses. Indeed, we are in the process of updating the

descriptions of our program (e.g., on our website) to make our mathematical requirements

more explicit.

Although not the primary rationale for revising our admission process, we note that our

approach is relevant to the question of the "shrinking pipeline in the STEM disciplines" [11-13].

As applied to mathematics and statistics, this pipeline begins in middle school (and before) with

students who are interested in mathematics. Some drop out of the pipeline in high school,

because math courses are uninteresting, irrelevant, poorly taught, and/or too hard for their

level of preparation. The same applies to their undergraduate experience, especially if they

encounter a "weed-out course", causing them to decide against a mathematically-related major.

Students with fewer math courses or having non-math majors are less likely to pursue graduate

study in a STEM field and ultimately receive a graduate degree. At each step, these challenges

disproportionately fall on students from disadvantaged groups, due to a combination of beliefs

(e.g., stereotyping), individual behavior, institutional behavior, resources (both formal and

informal) and inadequate support. The ultimate result is that too few students ultimately

succeed in the STEM fields, and also that the distribution of those who do is skewed toward

students from more advantaged backgrounds.

The points at which our program can intervene fall late in the process: (1) by broadening

admission requirements; (2) by enforcing consistency between admission requirements and

the curriculum; and (3) by striving to provide a supportive environment to all students,

recognizing that doing so requires due consideration of individual characteristics and social

context. Here, we have focused on the first two elements of the list: in essence, by returning

selected non-math majors who are traditionally assumed to have exited the pipeline back into

it, and by providing a curriculum for which they are adequately prepared. Broadening

admission requirements is especially reasonable for a "team science discipline" such as

biostatistics where a successful practitioner could, for example, build upon deep expertise in

biology and competence in mathematics rather than the reverse (which works well, too, of

course). Doing so would be less reasonable, for example, for those students who aspire to

doctoral study in theoretical mathematics, as this is a highly specialized discipline that directly

builds upon the depth of their previous mathematical training. We posit that an assumption

that is often unexamined is that biostatistics is a highly specialized and mathematically

intensive discipline rather like theoretical mathematics: this assumption holds true for a

minority of graduate programs (whose admission criteria should reflect this) but not for most

graduate programs and not for most practitioners.

Page 7 of 9

271

Troy, J., Grambow, S., Neely, M., Pomann, G.-M., Davenport, C., Ashner, M., & Samsa, G. (2024). Rationalizing the Mathematics Requirements for

a Master of Biostatistics Program: A Case Study and Commentary. Advances in Social Sciences Research Journal, 11(8). 265-273.

URL: http://dx.doi.org/10.14738/assrj.118.16706

As a practical matter, one recommendation we can provide to others is that implementing this

sort of change in admission requirements requires proper framing. In particular, the goal of

these changes was not to "admit students who are weaker at math", but instead to "enhance

consistency between what we teach and who we admit". For our program, who we admit

requires, at a minimum, acceptable background in each of the ABCs of biostatistics, with

mathematics being part of the analytic competency. Once they have entered the pool of those

with acceptable qualifications, applicants can distinguish themselves along multiple

dimensions. To those who might have worried that requirements are being weakened, we

clarified that (1) we continue to seek strong students, now with the criteria for what counts as

"strong" being broadened; and (2) every admitted student has sufficient background in

mathematics to succeed. Of course, we remain happy to admit students whose distinctive

strength falls within the domain of mathematics and whose exposure extends far beyond the

minimum.

CONCLUSION

In conclusion, we believe that removing unnecessary mathematical barriers and ensuring

alignment between our mathematics requirements, curriculum content, and admission

requirements is educationally sound and can also help increase access, reduce challenges, and

support success for aspiring or emerging biostatisticians from diverse or multifaceted

backgrounds. It is one way to contribute to a more inclusive and equitable learning

environment in our STEM discipline. We believe similar programs might benefit from

performing this type of analysis and reflection.

7 APPENDIX 1: ADMISSION CRITERIA FOR MATHEMATICS

In

Admission requirement Justification

Calculus

Good grades in math-related courses Some exposure to math is a surrogate for a generic ability

to define and solve statistical problems using math

Facility with basic mathematical concepts

such as functions

Used as a building block for what follows

Facility with integrals and derivatives for one

variable

Used for working with statistical distributions -- for

example, for transforming probability density functions to

cumulative distribution functions

Facility with integration in 2 variables

strongly preferred

Used for working with marginal and conditional

distributions

Ability to manipulate functions (e.g., to find

maxima and minima)

Used in working with likelihood functions, among others

Linear algebra

Ability to manipulate matrices (e.g., add,

invert, transpose)

Used as a building block for what follows

State models in matrix language Used to succinctly describe statistical models

Project high-dimensional data into lower

dimensions (e.g., eigenvalues, eigenvectors)

Used in multi-predictor models

Page 8 of 9

272

Advances in Social Sciences Research Journal (ASSRJ) Vol. 11, Issue 8, August-2024

Services for Science and Education – United Kingdom

8 APPENDIX 2: ILLUSTRATION OF FUNCTIONAL MASTERY OF CALCULUS

One of our admission criteria pertains to mastery of functions. To illustrate what is intended,

consider the following question:

"Is exp{-2t}, t>0, a function of t? How would you find its maximum value? Please

discuss "why" in addition to "how".

A possible answer is as follows:

A function is a rule, which takes inputs and uniquely assigns values for its outputs.

Here, the inputs are the positive numbers (i.e., "t>0"). For each value of t, the output

is f(t) = exp{-2t}. This is a special case where the usual calculus procedure for

finding a maximum of a function, which begins by finding the derivative of that

function and setting it equal to 0, doesn't work.

To see why the usual approach doesn't work, this derivative is (-2) *exp{-2t}.

According to the rules of exponents, exp{-2t} = 1 / exp{2t}, and there is no value of

x for which 1/x equals 0.

We can, nevertheless, proceed using logic. Simply plugging in values of t makes it

apparent that as t increases f(t) decreases. In essence, as t increases so does exp{2t}

and so 1 / exp{2t} decreases, as does 2 / exp{2t}. So, the maximum value of f(t)

corresponds to the smallest value of t within its range.

Now, if the range was t >= 0 rather than t > 0, we could plug t=0 into f(t) and obtain

the maximum value f(t)=1. As we allow t to decrease and approach 0, the value of

f(t) approaches 1 as closely as you like. Indeed, in the world of calculus, "approaches

1 as closely as you like" has the same interpretation as "is 1", and so the maximum

value of f(t) is 1.

Comment: To understand the principles behind the above solution, a working knowledge of the

nature of functions suggests that you should start by plotting f(t) on the y-axis and t on the x- axis and hope that this provides a clue about the location of the maximum. Plugging in a few

values of t makes it clear that f(t) decreases as t increases, and so the maximum must occur at

"t=0". This conclusion can be checked using rules of exponents and quotients, which are basic

mathematical techniques. Finally, the fundamental calculus principle of a limit translates

"approaches 1" to "equals 1".

The solution only relies upon basic mathematical manipulations, but does require the ability to

use of the construct of general mathematical thinking in order to set up the analytical approach.

In other words, general mathematical thinking suggests drawing a graph and using the shape

of that graph to discover the likely value of the maximum.

References

[1]. Biggs, J. Enhancing teaching through constructive alignment. High Educ 32, 347–364 (1996).

https://doi.org/10.1007/BF00138871

[2]. Neely ML, Troy JD, Gschwind GT, Pomann G-M, Grambow SC, Samsa GP. Preorientation curriculum: an

approach for preparing students with heterogeneous backgrounds for training in a Master of Biostatistics

Page 9 of 9

273

Troy, J., Grambow, S., Neely, M., Pomann, G.-M., Davenport, C., Ashner, M., & Samsa, G. (2024). Rationalizing the Mathematics Requirements for

a Master of Biostatistics Program: A Case Study and Commentary. Advances in Social Sciences Research Journal, 11(8). 265-273.

URL: http://dx.doi.org/10.14738/assrj.118.16706

program. Journal of Curriculum and Teaching. Online Published April 21, 2022. DOI:10.5430/jct.

v11n4p77.

[3]. Pomann G-M, Boulware LE, Chan CC, Grambow SC, Hanlon AL, Neely ML, Peskoe SB, Samsa G, Troy JD,

Yang LZ, Thomas SM. Experiential learning methods for biostatistics students: A model for embedding

student interns in academic medical centers. Stat. online 09Sep2022. https://doi.org/10.1002/sta4.506.

[4]. Samsa GP, LeBlanc TW, Locke SC, Troy JD, Pomann, GM. A model of cross-disciplinary communication for

collaborative statisticians: implications for curriculum design. 2018. Journal of Curriculum and Teaching.

Online August 3, 2018. DOI: 10.5430.jct.v7n2p1.

[5]. Samsa G. Using coding interviews as an organizational and evaluative framework for a graduate course in

programming. Journal of Curriculum and Teaching, 2020, 9(3), 107-140.

[6]. Samsa G. Evolution of a qualifying examination from a timed closed-book format to a collaborative take- home format: a case study and commentary. Journal of Curriculum and Teaching. 2021. 10(1). 47-55.

Doi:10.5430/jct. v10n1p47.

[7]. Troy JD, Neely ML, Grambow SC, Samsa GP. The Biomedical Research Pyramid: A model for the practice of

biostatistics. Journal of Curriculum and Teaching, 2021, 10(1), 10-17.

[8]. Troy JD, Granek J, Samsa GP, Pomann G-M, Updike S, Grambow SC, Neely ML. A course in biology and

communication skills for master of biostatistics students. Journal of Curriculum and Teaching. April 21,

2022. DOI: https://doi.org/10.5430/jct.v11n4p77.

[9]. Troy JD, McCormack K, Grambow SC, Pomann G-M, Samsa GP. Redesign of a first-year theory sequence in

biostatistics. Journal of Curriculum and Teaching. 2022;11(8),1-12. Doi:10.5430/jct.v11n8p1.

[10]. Troy JD, Pomann G-M, Neely ML, Grambow SC, Samsa GP. Are simulated coding interviews a fair and

practical examination format for non-professional programmers enrolled in a master's degree program in

biostatistics? Journal of Curriculum and Teaching. 2023, 11(15), DOI:10.5430/jct.v12n6p253.

[11]. Brown BA, Henderson JB, Gray S, Donovan B, Sullivan S, Patterson A, Waggstaff W. From description to

explanation: an empirical exploration of the African-American pipeline problem in STEM. Journal of

Research in Science Teaching, 2016, 53(1), 146-177.

[12]. Cannady MA, Greenwald E, Harris KN. Problematizing the STEM pipeline metaphor: is the STEM pipeline

metaphor serving our students and the STEM workforce? Science Education, 2014, 98(3), 443-460,

DOI:10.1102/sce.21108.

[13]. Estrada M, Burnett M, Campbell AG, Campbell PB, Denetclaw WF, Gutierrez CG, Hurtado S, John GH, Matsui

J, McGee RM, Okpodu CM, Robinson TJ, Summers MF, Werner-Washburne M, Zavala ME. Improving

underrepresented minority student persistence in STEM. CBE-Life Sciences Education. September 1, 2016

15:es5 DOI:10.10.1187/cbe.16-01-0038.