Report

Step Into the Role of Researcher Without Leaving the Classroom or Clinic Adirondack Assistive Technology Expo October 27, 2014 Rondalyn Whitney, PhD, OTR/L, FAOTA Clarkson University OT Department Acknowledgment for contributions from Sheila Braun and inspiration from Jason M. Molesky How to think like a researcher… What do you want to know? Let it Go! Where do you get ‘frozen’? Most of us get stuck on the analysis…. Image from https://www.google.com/search?rls=com.microsoft%3Aenus&rlz=1I7NDKB_enUS524&tbm=isch&q=frozen%20elsa&revid=1818079338ei =8ZU6VJ-KFsr8yQT07oKoAQ&ved=0CCMQsyU&biw=1024&bih=640 We use analysis all the time in practice…. How to think like a researcher… • What do you want to know? • Who else has investigated the topic? • How will you measure it? • How will you analyze what you measure? • What does it all mean? 1. Create your research question 2. Conduct a literature review 3. Establish the methodology 4. Determine the appropriate statistical measure for your question and run the analysis 5. Write the discussion – an interpretation of what you learned and how that compares to the existing literature. Basics of Analysis Method of analysis is predicted by your research question: Univariate analysis When you want to know the number/ count of data, often shown in the Mean score… Nominal data Frequency Results will report a count or a percentage Example: Number of boys and girls in a class. 20 boys, 40 girls (N = 60; boys n = 20, girls n = 40) 20/60 boys (33%); 40/60 girls (66%) Basics of Analysis Method of analysis is predicted by your research question: Bivariate analysis When you want to know if there is any correlation between two groups. The group of females and girl children have a correlation of 1, females and boy children have a correlation of 0. Therefore, correlation ranges between 0 and 1. The closer to 1, the higher the correlation. Correlation will give you an r value Example: On average scores for difficult child behaviors are strongly correlated with maternal stress (r =.821; p = .001). Basics of Analysis Method of analysis is predicted by your research question: Bivariate analysis When you want to know if there is a difference between two groups. Did the intervention provided change the speed of the child’s performance on written tasks? Did the intervention provided improve the speed of the child’s performance on written tasks? A t test measures the difference between two groups. One-tailed measures direction, two tailed measures change. Example1: Using a one-tailed test, the control group did not show lower stress than the experimental group (p=.045). No relationship was found between high levels of socially disruptive child behaviors and mother’s perceived support in the community (p=.798) but both family support (r =.267, p=.005) and feelings of cherishment (r =.216, p=.025) showed a weak correlation. p value reports the t-test r value reports the correlation Most clinical research questions can be answered with univariate or bivariate analysis Basics of Analysis Method of analysis is predicted by your research question: Multivariate analysis Example: When you want to know how one variable changes as a result of the other. Negative emotion laden words accounts for 17.6% of the total variance in total stress (F(1, 34) = 7.252, B = -14.894, p = .011). Question: What is the strongest predictor of maternal stress when raising a child with disruptive behaviors? The greater the number of negative emotion laden words, the lower the total stress. Results 3: Multivariate 1. Negative emotion laden words account for 22.8% of the variance in parental distress (not controlling for anything else; F(1, 34) = 10.060, B = 6.948, p = .003). 2. Negative emotion laden words also account for 17.6% of the total variance in total stress (F(1, 34) = 7.252, B = -14.894, p = .011). The greater the number of negative emotion laden words, the lower the total stress. 3. A strong and significant correlation was found between Difficult Child behaviors and maternal stress, both before (r =.810; p= .001) and after the intervention (r = .891; p=.001). 4. Perceived support predicts maternal stress and Quality of Mother-Child Relationship: as stress increased , support increased [community (p =. 005), family (p = .003), and cherishment (p = .010)]: As dysfunctional relationship increased, so did support[community (p = .003), cherishment (p = .045), Family not a significant predictor (p = .516)] 5. ASD droped out as a predictor – cherishment and SPD held in almost to the end but family support highest (B = 1.98, p = .001), difficult child and dysfunctional mother-child relationship were covariates. *bivariate used in regression model 12 Final Model: Average Total Stress 1.228 Average ParentChild Interaction 1.171 1.987 Average Difficult Child Score 100% of the Variance in Average Total Stress Family Support Plus a mix of things we don’t know about (12.9% of it) 13 All I need to know about research I learned from a bag of M&Ms… What questions can we ask of a bag of M&Ms What kind of variable represents your M&Ms most usefully? A.Binary? B. Continuous? C. Categorical? (a.k.a. “nominal”) Huh? Are the M&Ms binary? • You have a binary sample if you have things that can be in one state or another. Binary means 2, like yes or no o Example: M&Ms might be eaten or not eaten. o We can call “eaten” a 0 and “not eaten” a 1 (which gives us a value) or leave as a nominal variable (i.e. a name). • You should have a sample that are all “not eaten” (presumably). • This is a CONSTANT. All the values are exactly the same for all the M&Ms. Continuous? • A continuous variable can hold any value • You have a certain number of M&Ms, so you might say that “M&M” is a continuous variable. • You have only 1 bag of M&Ms • If you had 100 bags of M&Ms of different sizes you might say the number of M&Ms into your data set for each bag “One bag of multicolored M&Ms (N = 42)” • Using a continuous variable, you could come up with the “mean value” or “average” number of M&Ms found in each bag Is the variable “M&M” categorical? • You have a categorical variable if you have a bunch of states (in the case of M&Ms the states are colors, such as red, yellow, green brown, & blue). • Then you can set each CATEGORY to a number that represents how many you have. This is the most useful unit of analysis for this study! Break into small groups 1. separate your M&Ms into colors 2. Create an analysis of your data 3. Report on your findings when we return as a large group results The sample of M&Ms, (N = 57) included a six colors: orange, brown, green, yellow, blue and red. Orange and brown appeared with greatest frequency in the sample (n = 12). The sample distribution is shown in Table 1. M&M distribution (example) Blue Orange Brown Red Green Yellow 8 12 12 6 10 9 Results Make sure your answers (results) answer your question. • Did you ask how many? Results should give a count of the total sample • Did you ask how many of each? Results should give a count of the sub-variables • Did you ask about the mean differences between two things? Then you should be comparing the means (averages) ……Etcetera Etcetera Etcetera …. What’s the best way to tell people about it VISUALLY? OR Bar chart? Pie chart? Either way. BUT… • APA 6th recommends using a pie chart to show how a TOTAL POPULATION is broken down into categories. Here our total population is the population of M&Ms in your bag compared with those in the bags of your classmates. M&Ms per Bag 150 100 50 0 Bag 1 Bag 2 Bag 3 • Use a bar chart when you are showing just a portion of the population or a series of scalar variables. So which is better?. A.Pie chart B.Bar chart Table 1: Color distribution in Sample of M&Ms Yellow Green Blue Red Orange Brown Red Green Brown Yellow Orange Blue 0 2 4 6 8 10 12 14 12 10 Blue 8 Orange Brown Red 6 Green Yellow 4 2 0 Blue Orange Brown Red Green Yellow T1: COLOR DISTRIBUTION (3 GROUPS) Grand Total Yellow Brown Grand Total G3 Blue G2 G1 Green Red Orange 0 20 40 60 80 100 120 140 160 Figure 2. Trial 1 bar graph of color distribution for all groups (N = 3) 180 DISTRIBUTION (ALL GROUPS) Grand Total Yellow Brown Grand Total G3 Blue G2 G1 Green Red Orange 0 20 40 60 80 100 120 140 Figure 6. Trial 2 bar graph of color distribution for all groups 160 180 Who cares about M&M studies? • Connecting the dots to practice…. • Nominal is nominal is nominal • Categorical is categorical is categorical… Image from http://giantgag.net/chocolate-is-the-answer-who-careswhat-the-question-is/ Important Academic Verbiage • Show and report categorical variables or binary variables using “frequencies” of each category (“How frequently does it occur?”). • Show scalar variables using their “mean” and “standard deviation.” Example of Academic Verbiage • “The sample of M&Ms (N = 22) consisted of six colors: red (n = 1), brown (n = 2), green (n = 3), orange (n = 4), yellow (n = 5), and blue (n = 7).” • Note the use of past tense. This is correct APA language. • Note the use of a capital N to represent the entire sample and the use of lowercase ns to represent the individual groups. • Note the use of italics to show that you are giving “a statistic,” or a number that describes your sample. Another APA Rule • Never, ever, EVER use BOTH words AND pictures to discuss your sample. Use either words or an image. • For instance, to use only the image, write, “See Figure 1 for the breakdown of frequencies in the sample.” • Do not repeat yourself. • Do not repeat yourself! Figure 1. Pie chart of M&Ms. References Molesky, J.M. (n.d.) Everything I Ever Needed to Learn about AP Statistics I Learned From a Bag of m&m’s: Class Activities for Advanced Placement Statistics http://web.mac.com/statsmonkey