4 hypotheses complex relationships and - lippincott...

Supplement to Chapter 4 Complex Relationships and Hypotheses • 4-1

4-1

As discussed in the textbook, some studies focus on research questions that are fairly

straightforward and that involve understanding the relationship between two variables. Examples may include exploring:

• The effect of an illness (the independent variable or IV) on fatigue (the dependent variable or DV),

• The effect of a nursing intervention (IV) on pain (DV), or

• The effect of pain (IV) on quality of life (DV).

Sometimes, however, researchers study multiple variables that may be interrelated in complex ways.

This Supplement describes some aspects of complex relationships and complex hypotheses. We begin by explaining moderating and mediat-ing variables. For those who wish to learn more, Bennett (2000) has written a useful description of the conceptual and statistical differences between the two.

MODERATING VARIABLES

A moderator variable (MV) affects the strength or direction of an association between the indepen-dent and dependent variable. Identifying modera-tors may be important in understanding when and for whom to expect a relationship between the IV and DV, and it often has clinical relevance—for

example, who benefits the most or the least from an intervention.

Research Question 1: What is the effect of nurses’ use of humor (versus the absence of humor, the IV) on stress (the DV) in hospital-ized cancer patients (the population)?

Research Question 2: Does nurses’ use of humor have a different effect on stress in male versus female patients?

Research Question 1 is a simple research question with only two variables. Research Question 2 ex-amines whether the relationship between the IV and the DV is influenced by or moderated by a third variable. In this example, gender is an MV. Moderator (or moderating) variables can be charac-teristics of the population (e.g., male versus female patients), of the circumstances (e.g., rural versus urban settings), or of external agents (e.g., male versus female nurses using humor). The follow-ing are examples of question templates that involve an MV:

• Treatment, intervention: In (population), does the effect of (IV: intervention) on (DV) vary by (MV)?

• Prognosis: In (population), does the effect of (IV: disease, condition) on (DV) vary by (MV)?

• Etiology, harm: In (population), does (IV: ex-posure, characteristic) cause or increase risk of (DV) differentially by (MV)?

Complex Relationships and Hypotheses4

Polit10e_Supp_Ch04.indd 4-1Polit10e_Supp_Ch04.indd 4-1 3/10/16 3:39 AM3/10/16 3:39 AM

4-2 • Part 2 Conceptualizing and Planning a Study to Generate Evidence for Nursing

SIMPLE AND COMPLEX HYPOTHESES

As mentioned in the textbook, hypotheses can be either simple or complex. We define a simple hypothesis as a hypothesis that states the predicted relationship be-tween a single IV and a single DV.

Simple HypothesesFigure 4.1 presents a diagram illustrating a simple hypothesis. In this figure, the hypothesis predicts a relationship between one IV and one DV (outcome). In Figure 4.1, the IV is the presumed cause, influ-ence, or antecedent of DV, which is the predicted outcome, effect, or consequence. The circles repre-sent IV and DV, and the hatched area where the two circles overlap designates the strength of the relation-ship between them. If the variables were unrelated, the circles would not overlap at all. For example, let us suppose that the IV is smoking (versus not smok-ing) and the DV is having (versus not having) lung cancer. If the circles did not overlap, it would suggest that smoking (IV) was unrelated to lung cancer. If the two circles overlapped completely, it would sug-gest that everyone who smoked developed lung can-cer. The relationship depicted in Figure 4.1 suggests that smoking status is related to lung cancer but that some smokers do not get cancer—and some people with lung cancer were not smokers.

Example of a Simple Hypothesis: Nafiu and Onyewuche (2014) hypothesized that abdominal obesity (IV) increases the risk of perioperative ad-verse respiratory events (DV) in children undergoing elective, noncardiac surgeries.

MEDIATING VARIABLES

When a study purpose is to understand causal path-ways, research questions may involve a mediating variable—a variable that intervenes between the IV and the DV and helps to explain why the relation-ship exists. For example, we might ask the follow-ing: Does nurses’ use of humor have a direct effect on the stress of hospitalized patients with cancer, or is the effect mediated by humor’s effect on immune function (natural killer cell activity)? A mediating variable can be conceptualized like this:

humor → natural killer cell activity → patient stress

This means that the path through which nurses’ humor affects stress is its effect on natural killer cell activity.

Many of the theoretical models described in Chapter 6 of the book involve mediating factors. Important health outcomes are often not directly affected by nursing actions but rather by their ef-fects on such factors as self-efficacy, anxiety, im-proved health-promoting behaviors, and so on. In intervention research, moderating variables help researchers better understand how the intervention works. Chapter 18 of the book provides an exam-ple of how hypotheses about mediating variables are tested.

T IP : For those with strong statistical skills, a paper by Levy and colleagues (2011) discusses advances in statistical methods for testing hypotheses about mediation effects. This paper is available online as an open-access article, and a link to it is provided in the Toolkit on for Chapter 4.

Example of Moderating and Mediating Variables: Christopherson and Conner (2012) studied health-risk behaviors in late adolescence. In their study of 437 adolescents, loneliness was a me-diating variable, mediating the relationship between parental attachment and smoking. Gender moder-ated the relationships, and so separate analyses were undertaken for males and females.

FIGURE 4.1 Schematic representation of a simple hypothesis. IV, independent variable; DV, dependent variable.

IV DV


Supplement to Chapter 4 Complex Relationships and Hypotheses • 4-3

Complex HypothesesMost phenomena are affected by a multiplicity of factors. A person’s weight, for example, is af-fected simultaneously by such factors as height, diet, bone structure, activity level, and metabolism. If the DV were weight and the IV was a person’s caloric intake, we would not be able to explain or understand individual variation in weight very well. For example, knowing that Alex O’Hara’s daily ca-loric intake averages 2,500 calories would not per-mit a good prediction of his weight. The overlap in circles indicating the strength of the relationship between caloric intake and weight would likely be smaller than what is shown in Figure 4.1.

Many other factors are related to a person’s weight, however. Knowledge of those factors, such as Alex’s height, would improve our ability to accu-rately understand and predict his weight. Figure 4.2 presents a schematic representation of a complex hypothesis, showing the situation in which the DV is influenced by two IVs (IV1 and IV2). To pursue the preceding example, the hypothesis might be: Taller people (IV1) and people with higher caloric intake (IV2) weigh more (DV) than shorter people and those with lower caloric intake. In this example, we expect that caloric intake and height would do a better job in helping us explain variation in weight (DV) than caloric intake alone. Complex hypoth-eses have the advantage of allowing researchers to capture some of the complexity of the real world.

Example of a Complex Hypothesis with Multiple Independent Variables: Jolley and col-leagues (2014) hypothesized that age (IV1), illness severity (IV2), and presence of a tracheostomy (IV3) would predict levels of the receipt of a physical therapy consultation among patients requiring pro-longed mechanical ventilation.

Just as a phenomenon can be caused or influ-enced by more than one IV, a single IV can also influence more than one phenomenon, as illustrated in Figure 4.3. A number of studies have found, for example, that cigarette smoking (IV) can lead to both lung cancer (DV1) and coronary disorders (DV2). Complex hypotheses are common in studies that try to assess the effect of a nursing intervention on multiple outcomes.

Example of a Complex Hypothesis—Multiple Dependent Variables: Grey and colleagues (2011) hypothesized that participation in a coping skills training program for parents of children with type 1 diabetes, compared to nonparticipation (IV), would result in better coping (DV1), quality of life (DV2), and children’s metabolic control (DV3).

Although hypotheses can be even more complex (e.g., two IVs predicting two DVs), the hypotheses in nursing studies often are like the ones shown in Figures 4.1, 4.2, or 4.3.

FIGURE 4.2 Schematic representation of a complex hypothesis. IV, independent variable; DV, dependent variable.

IV1

IV2

DV

FIGURE 4.3 Schematic representation of a complex hypothesis. IV, independent variable; DV, dependent variable.

IV

DV1

DV2


4-4 • Part 2 Conceptualizing and Planning a Study to Generate Evidence for Nursing

*Grey, M., Jaser, S., Whittemore, R., Jeon, S., & Lindemann, E. (2011). Coping skills training for parents of children with type 1 diabetes: 12-Month outcomes. Nursing Research, 60, 173–181.

Jolley, S. E., Caldwell, E., & Hough, C. (2014). Factors associ-ated with receipt of physical therapy consultation in patients requiring prolonged mechanical ventilation. Dimensions of Critical Care Nursing, 33, 160–167.

*Levy, J.A., Lenderman, L., & Davis, L. (2011). Advances in mediation analysis can facilitate nursing research. Nursing Research, 60, 333–339.

Nafiu, O. O., & Onyewuche, V. (2014). Association of abdomi-nal obesity in children with perioperative respiratory adverse events. Journal of Perianesthesia Nursing, 29, 84–93.

Sawatzky, R., Ratner, P., Richardson, C., Washburn, C., Sudmant, W., & Mirwaldt, P. (2012). Stress and depression in students: The mediating role of stress management self-efficacy. Nursing Research, 61, 13–21.

*A link to this open-access journal article is pro-vided in the Toolkit on for this chapter in the accompanying Resource Manual.

Hypotheses are also complex if mediating or mod-erator variables are included in the prediction. For example, it might be hypothesized that the effect of caloric intake (X) on weight (Y) is moderated by gen-der (Z)—that is, the relationship between height and weight is different for men and women. Or, we might predict that the effect of ephedra (X) on weight (Y) is in-direct, mediated by ephedra’s effect on metabolism (Z).

REFERENCES CITED IN CHAPTER 4 SUPPLEMENT

Bennett, J. A. (2000). Mediator and moderator variables in nursing research: Conceptual and statistical differences. Research in Nursing & Health, 23, 415–420.

Christopherson, T. M., & Conner, B. T. (2012). Mediation of late adolescent health-risk behaviors and gender influences. Public Health Nursing, 29, 510–524.


4 hypotheses complex relationships and - lippincott...

Documents