GOT DATA? Step-by-Step Guide to Making Data Work for You Center for Applied Research Solutions, Inc 771 Oak Avenue Parkway, Suite 3 Folsom, CA 95630 (916) 983 -9506 TEL (916) 983 -5738 FAX
GOT DATA? Step-by-Step Guide to Making Data Work for You Facilitators: Kerrilyn Scott Christina Borbely Produced and Conducted by the Center for Applied Research Solutions, Inc. for the California Department of Alcohol and Drug Programs SDFSC Workshop-by-Request May 16, 2005 Authored by Christina J. Borbely, Ph. D. Safe and Drug Free Schools and Communities Technical Assistance Project
Objectives l l l Preparing to Use Data – Database options & structure – Identifying data – Coding & Entering – Storing & Cleaning Methods for Summarizing Data – Basics: frequency & % change – Beyond Basics: mean scores; making comparisons Interpreting Data Effective Report Writing Utilizing & Disseminating Findings – Program improvement, Funders, Key Stakeholders
Ready, Set, Go! Preparing to Use Data Ø Ø Database Options Identifying Data Coding Data Entering, Storing, & Cleaning Data
Database Options l l l Microsoft Excel Microsoft Access SPSS
Excel l l Spreadsheet format Some computational functions Compatible with other MS software & statistical software Comes with Microsoft Office package (or $299) http: //office. microsoft. com/enus/FX 010858001033. aspx
Access l l l User friendly design Requires some preparation prior to data entry Generates custom reports Good for qualitative (i. e. open-ended items) & quantitative data Compatible with other Microsoftware & statistical software (i. e. converts easily to Excel!) Comes with Microsoft Office package (or $299) http: //office. microsoft. com/en-us/FX 010857911033. aspx
SPSS l l Spreadsheet format Requires some tutorial (not always intuitive) One-touch data analysis! Pricing ranges from $599 to $1499 www. spss. com
Watcha Got? l l Identifying data Variable names
Identifying Data l Each piece of information you have for a participant or a program is data. Data are… # of completed surveys data # of times a youth attended a session # of youth who attended a meeting # of merchants contacted for outreach Age Grade
FYI: Types of Data l Discrete, categorical Male/Female US Citizen/Non US Citizen Freshman, Sophmore, Junior, Senior l Continuous Age Salary Conflict Resolution Ability
Variable Names l Each piece of data is labeled with a unique (and hopefully meaningful) variable name. Data Section E, item 3 Age Unit 1 total score Variable Name E 3 Age Un 1 tot
Variable Names: Do’s & Don’ts l Meaningful For section E, item 6: E 6 l DOB E 6 Variable 124 a Short Date of Birth Youth Survey Section E, Item #6 l E 6, E 7, E 9, F 1, F 2 Systematic 1 F, two. F, Fthree
Plan to reference data collection time points First administration: l BL (for baseline) or T 1 (for time 1) or PRE (for pre-test) BLE 6, FUE 6 l E 6, E 6 Be consistent with the chosen system T 1 E 6, T 2 E 6 E 6 T 1, T 2 E 6
Coding Key: Do’s l Translate into numeric values For response scale: YES! Yes No NO! YES! = 3 Yes = 2 No = 1 NO! = 0 l Record coding key directly onto measure & save!
Coding System Examples Race Black = 1 Hispanic = 2 White = 3 Asian = 4 Other = 5 Gender Male = 1 Female = 2
Coding Key: Don’ts l Do not create a separate variable to code each response to an item. 1. What grade are you in? A. 6 th B. 7 th C. 8 th Variable name = BL 1 Codes A=1; B=2; C=3 NOT Variable name = BL 1 A; BL 1 B; BL 1 C Codes Yes=1; No=0
Advanced Coding • Collapsing Variables by Code 1. Do you live in a house? Y/N 2. Do you live in an apartment? Y/N 3. Do you live in a barn? Y/N Variable Name: “Reside” Codes: house = 1 apartment = 2 barn = 3
l. Reverse Coding The values of the coding system may need to be reversed to reflect the true meaning of the response. 1. Do you runaway from home? Often Sometimes Rarely Never 2. Do your parents smile at you? Often Sometimes Rarely Never 3. Are you happy at home? Often Sometimes Rarely Never Variable codes: 4 3 2 1 Reverse code: 1 2 3 4
Entering Data in Your Database l Create 1 row of variable names: Across l Create 1 column of names/id #s: Down l Enter post test & follow-ups by extending the row for each participant ID BLgrade BLa 23 0025 6 2. 5 l T 2 grade T 2 a 23 7 3. 1 Save regularly as you enter (don’t lose all that work!)
Storing Data l l Hardcopies Electronic files
Under Lock n’ Key l Guard with your life until a back up is made l Keep all hardcopies as backup l Maintain back ups in different locations l Preserve confidentiality l l Separate identifying information from surveys Use passwords; locked file cabinets; secured offices
Cleaning Data: Quick, Easy, & Worth It! Save yourself the grief of inexplicable scores… l Ø Ø Data should fall within an expected range (e. g. 1 to 5). Scan data for unusual numbers by: Visual review A “sort by” function A “find” function A “minimum/maximum” or “range” function
Squeaky Clean! l Use a “missing” marker (e. g. 999) when a response is purposely missing (e. g. left blank, etc. ) Pros: easy to spot unintentionally unentered data Cons: extra step to remove missing marker later l Don’t forget to exclude “missing” data values, so it doesn’t mess up your computations!
FYI: How to use “missing” markers 1. 2. 3. 4. 5. Select number or symbol that will not naturally occur in the data Enter marker when data point is unavailable Clean data & look for “blanks”. Fill in unentered or incomplete data. After data is clean, delete or exclude the missing marker Do data analysis
Recommendations l Consider using “in house” resources for entering & cleaning data l Consider outsourcing database development to a graduate student or local evaluator
FYI: Outliers l An outlier is a data point that does not cluster with other data points in the group. l Example: ages range from 12. 1 to 14. 3 years; there are 3 outliers age 17. 4; 19. 2; and 19. 7 years. l It may skew data so that it is not representative of the sample. l Consider excluding outliers
Housekeeping Activity: Clean the Data
Guide: Step 1 ü Set up a database ü Code and enter data ü Clean database Kids today!
Putting Data to Work Methods for Summarizing Data ØBasics ØTaking It Up a Notch
Add It Up l Count or Tally Do you attend Club Live? Yes No No ll Blank l By computer T 21 a jn 789 1 rs 587 0 ty 390 By hand llll Yes ID 1 ge 188 1 bo 989 0 pc 490 1 sz 688 1 Yes=1; No=0; Blank=999 va 689 999
Frequencies: Ratio & Percent Distribution Quantifies rate of occurrence for categories of information Useful for…. What race are you? Do you live with both q Black biological parents? q White Yes No q Asian q Hispanic q Other NOT As Useful for…. How much you like school? (circle one) YES! Yes No NO! How old are you? _____
Calculating Frequency l Sum the number of times a given response occurs l Report a number: a ratio or percentage Gender Male Female Total Ø Ø # of participants 49 51 100 % of participants 49% 51% 100% Of the 100 participants, 49 were male. This year, almost half (51%) of the participants were females.
Common uses l Demographics to characterize participants or community Race; gender; grade; homeowner status l Statistics l Impact to describe program Number of program completers % of city council members contacted statements on outcomes % of youth reporting ATOD use Ratio of signage below adult eye-level
Reporting Frequencies Frequency of participants reporting they are: Male Employed Getting mostly B’s in math Parents of a FNL youth Frequency with which: Decoy buys are successful Alcohol-sponsored events occur
Sample: Excerpt of Frequency in Text “Of clients with completed CBCL/YSR, well over half (56. 9%) function in the lowest quartile of global competence. Specifically, clients demonstrate compromised ability related to engagement in ageappropriate activities, social interaction, and performance at school. Given that services are provided in the school context, it is not surprising that almost three-quarters of the clients (71. 2%) function in the bottom quartile of school-related competence. Teachers and other school staff, individuals familiar with indicators of school competence, are the most common referral source of students. It is expected that competence in these domains will benefit from student participation in counseling services. Additional data is being collected to test for improvement over time. ”
Change Score l Comparison of scores to assess change Proposed outcome: 80% of youth increased awareness of ATOD consequences ID T 1 consq T 2 consq Change Increase jn 789 3. 4 0 No rs 587 2. 1 3. 6 +1. 5 Yes ty 390 2. 5 3. 4 +. 9 Yes ge 188 3. 0 3. 5 +. 5 Yes bo 989 4. 3 4. 5 +. 2 Yes va 689 999 2. 9 999 N/A pc 490 3. 2 2. 9 -. 3 No sz 688 1. 6 2. 5 +. 9 Yes 5 of 7 youth increased scores = 71. 4% of youth increased awareness of ATOD consequences
Taking It Up a Notch l l Mean scores And beyond…
Mean Scores l The mean refers to a variables central tendency and is the sum of all a factors values divided by the number of values. “Mean” and “average” refer to the same concept. l
Calculating Means l Sum all the response values, then divide by the total number (of responses or items) l Provide a frame of reference (“out of how many”)
Averages ID aj 785 Age 20 Item. E 7 4 Rsk. Fctrs 3 tk 983 22 3 0 mr 286 19 61/3 = 20. 3 5 12/3 = 4 2 5/3 = 1. 6 Ø Ø Ø The mean age of the participants is 20. 3 years. The average score on Item E 7 is 4 out of 5. Youth have an average of 1. 6 risk factors out of a possible 4 risk factors.
Common Uses l To make a generalized statement about a group. l Demographics to characterize participants or community Age; Income level l Impact statements on outcomes Level of ATOD use among youth Sub-scale scores
Reporting Mean Scores l Report means of sub-scales Average score for “Community Connection” scale l Report mean scores of an individual item Item E 4: How often did you smoke pot in the past 7 days? l Report mean score of occurrence Average number of hours spent educating merchants
Sample: Excerpt of Mean Score in Text “Of the districts completing Year 1 Superintendent Surveys, the majority indicated that counseling services were of a resource of high value. On a five-point scale with 5 being the highest value, the average value assigned to the Project X counseling services was 3. 67. In addition, all districts indicated that parents, teachers, administrators, and school psychologists were largely receptive to and supportive of the resource. The majority of responding superintendents indicate that districts would benefit from expanding counseling services and improving the physical space allotted for service delivery. Clearly, Year 1 has culminated in substantiated need and the resolve to prioritize addressing the need. “
Analysis Activity: Findings
FYI: Calculating Subscale Means For each case, sum the values for all items in the subscale and divide by that number of items. Then calculate the overall mean of each participant’s mean score. l l Subscale: Attitudes Towards Violence Items included: A 8, A 9, B 4, E 7, F 2, F 3 (6 items) ID A 8 N 7 H 2 K 2 F 1 A 9 3 2 B 4 5 2 E 7 1 1 F 2 F 3 Sum Mean 3 3 17 2. 83 4 1 11 1. 83 Overall mean: 4. 66/2 Attitudes Toward Violence mean = 2. 33 out of 5.
FYI: Other Measures of Central Tendency l. Mode The most frequently occurring value in a set of values The modal response for the smoking subscale was 2. 0 out of 5. 0. This indicates that while youth may have tried smoking, most do not smoke on a regular basis. l. Median It is the value that is the mid-point in a set of values where half the values are smaller & half are larger. The median cost of a home in the area is $350, 000, well above the average family income for participating parents.
Apples to Apples? Comparing Frequencies & Means to Means and Frequencies to Frequencies l Over time Pre- to Post-Test Scores Incidence statistics the year before to the year after the program l Across groups Program participants to control/comparison group Merchants with low program participation to those with high program participation
Frequencies & Means: Method of Comparison l Eyeballing differences l Testing differences Ø Anyone can do it Limits interpretation Ø Requires a simple statistical test Determines whether the difference is meaningful Allows definitive statement about comparison Ø Ø Ø
Recommendations l Consider using an evaluator or identifying a consultant (like a local graduate student) to do statistical tests or analyses.
Activity: Compare Scores
And beyond… l l l Normal distribution Standard deviation Statistical significance
FYI: Normal Distribution l Normal distribution refers to a group of data points that occur symmetrically and with a bell-shaped density and one peak. Balasubramanian Narasimhan , Stanford University, July 22, 1996
FYI: Standard Deviation Standard deviation is a standardized score to indicate where a finding falls on the normal distribution. l Often means are reported with a standard deviation (SD). For example, mean = 3. 4 (. 17). l l Rule of thumb: SD between 0 and 2 are fine. Outside of this, finding may be skewed.
FYI: Statistical Significance l l l “Statistical significance” refers to the probability that the outcome of data analysis indicates an effect when there isn’t one. When comparing means or frequencies (or other analysis outcomes), a test statistic is used to determine if there is a meaningful difference. If a finding is “significant”, the outcome is considered true (with 95% certainty)
Guide: Step 2 ü Compute frequencies & means to describe program, participants, and outcomes ü Compare findings
Making Meaning Interpreting Data ØBite your tongue ØSound bytes ØSpin
Bite your tongue Do say Is associated with… This suggests… May indicate… Appears to… *note specific limitations (e. g. “no baseline”) Don’t say Is caused by… Is the result of… Due to… Because of… Significantly differed… (unless tested)
Sound Bites l Put favorable findings in short, sweet sentences. Statement to press: Participation in FNL is associated with increases in selfesteem. Presentation to key stakeholders: Merchants with Responsible Merchant education are less likely to sell alcohol to children in our community.
Spin l Group your data to maximize findings (e. g. cut a 4 point scale into high/low scores) l “No change” is “maintenance” l Couch unanticipated or lack of findings in reasonable explanation or plans
Sample: Excerpt of “Spin” “While scores from the Youth Survey’s ATOD use subscale indicate that participants increased their substance use over time (mean use at baseline = 3. 2 out of 5; post-test = 4. 0), this should be considered in context. Specifically, research reports a developmental (i. e. maturational) effect on adolescent use rates. When participant use rates were compared to rates reported for the same age group in Sutter County’s California Healthy Kids Survey, our youth were faring better…. . In addition, as noted in the previous section, youth attitudes towards substance use showed improvement over time. Within the literature, change in attitude is widely regarded as the first step in impacting behavior…”
Try Your Hand Activity: Translating Findings in Text
Guide: Step 3 ü Find your findings ü Select key findings ü Strategically frame findings
Read All About It! Effective Report Writing ØKnow Your Audience ØShow Your Work ØIf You’ve Got It, Flaunt It ØA Picture Speaks a Thousand Words
Know Your Audience l Consider Who You Are Dealing With…. What do they care about? How much time do they have? What level of detail is ideal? What do you expect to accomplish by sharing information with them?
Show Your Work l Don’t just report findings, report how you found them Specify: measure development administration details
Sample: Excerpt of Evaluation Methodology “The evaluation components for the fifth year consisted of the following elements: Protégé Pre- and Post Surveys (Outcome), Participant Satisfaction Surveys (for both Mentors and Proteges), Program Advisor Surveys, County Coordinator Phone Interviews, and Site Visits (Primarily focusing on interviewing the youth participants). In addition, Monthly Reports and Mentoring Session Activity Logs were collected analyzed. The following provides a more detailed description of each component and the sample size. Protégé Pre- and Post-Surveys: These instruments are intended to measure the impact of FNLM on the Proteges. The survey measures constructs such as school attachment, decision making and goal setting, conflict resolution, refusal skills, and ATOD harm perceptions and use. The survey took approximately 30 minutes to complete and primarily consisted of likert scale rating items. A subsample of seven counties were administered the outcome survey and a total of 71 matched pre/post surveys were used for the data analysis” (FNL Year 5 Evaluation Report, 2003)
l Describe program or evaluation “lessons learned” to account for modifications Sample: Excerpt of Limitations “Over the course of the four years, tracking program attendance has posed a major challenge. During the first two years of the program, valid attendance rates were not available due to hand-written attendance logs and a lack of documentation of program drop-outs. While the attendance data has improved dramatically over the course of the program, for the previous year tracking individual attendance rates was still not feasible. However, appropriate and consistent documentation of the number of mentors and protégés attending the mentoring sessions did allow for reporting average attendance levels by school site for mentors and protégés for each county; this method will be continued this year and will further benefit from the addition of an attendance summary sheet. ” (FNL Year 5 Evaluation Report, 2003)
If You’ve Got It, Flaunt It l Insert statistics to describe the program, staff, and participants l Report overall outcome findings & include notable specifics “At program end, fewer youth showed favorable attitudes toward alcohol use (17% versus 25%), especially girls (13% versus 29%). ”
A picture speaks a thousand words l Use graphs & charts to illustrate findings FNL survey relationship building mean scores (N = 204 youth) Percentage of Participants at Risk (N = 100) Community ATOD indicators per 1, 000 population
When To Use What Graphic l Line Graph: use to display values (data points) over time Community ATOD indicators per 1, 000 population l Bar Chart: use to display a distribution of values across categories
l Grouped Bar Chart: use to display a distribution of values across categories for two+ variables l Pie Chart: use to display the distribution of cases across categories. Wedge=number or percentage. FNL survey relationship building mean scores (N = 204 youth) Percentage of Participants at Risk (N = 100)
Creating Figures l l l Automated “chart” function in MS Word (2003), Excel, & Powerpoint Label everything Give each figure an informative title “Mean survey scores of 10 th grade Youth Coalition members at Oak Ridge High School” l l Give context of data (e. g. “per 1000 population”) Indicate the population size (e. g. “ 112 participants”)
Writing About Graphics Do highlight key findings displayed in figure Don’t reiterate in text every detail of the figure Sample: Excerpt of Text for Figure “Retention Rates: The attrition rates continue to show a slight decrease this year. A total of Fifty-two mentors (10%) and thirty-eight protégés (9%) were reported to drop-out of the program. In the previous year, attrition rates of 13% were seen for both the mentors (n=147) and proteges (n=141). The average number of protégés that dropped from all school sites was higher than for the mentors. “ (FNL Year 5 Evaluation Report, 2003)
Writing Tips Say it in numbers The letter-writing campaign was successful. The letter-writing campaign resulted in communication with 67 local government officials. Say ONE thing at a time More than half of the original participants completed the program and relationships with parents improved over time. More than half (57%) of the original participants completed the program. These young people demonstrated a 12% improvement in relationships with parents over time.
Writing Tips l Be precise (not vague) Program participants included high risk youth. Over one-third (36%) of program participants met at least one of three risk factors, including school expulsion/drop out, juvenile arrest record, or free-lunch status. l Connect proposed outcomes to performance measures to findings. A primary goal of the program was to reduce accessibility of tobacco to minors. Decoy buy assessments and focus groups with merchants involved in the merchant education program indicate that tobacco accessibility is more stringent now compared to at Year 1.
Writing Tips l Add interpretation or explanation to outcomes. Results from the Youth Survey indicate that teens showed healthier attitudes toward drug use, but increased drug use behavior over time. It may be that the program is most effective in impacting youth attitudes, not behaviors related to drug use. Research suggests that appropriate attitudes is a first step towards changing behavior.
Writing Tips l. Use qualitative data to add depth to quantitative data. Program records indicate that after a mid-year dip in attendance rates, regular participation exceeded expectations. Program records indicate that after a mid-year dip in attendance rates, regular participation exceeded expectations. A focus group conducted with program staff at the end of the school year revealed that a gang violence incident on campus resulted in the temporary suspension of all after school activity programs. This corresponds with the dip in our program’s attendance rates.
Try Your Hand Activity: Writing Up Findings
Guide: Step 4 ü ü ü Describe how evaluation was conducted Include general and detailed findings Consider using graphics
Spread the Word Utilizing & Disseminating Findings
One-pager l Develop a 1 page summary to: – – – describe program key impacts recommendations or next steps
Sample: Excerpt of One Page Summary
Program Improvement l l Identify strengths and weaknesses Use findings to inform strategic planning Regularly report impact to project staff to for morale boosting sessions Highlight modifications made based on lessons learned
Sample: Excerpt of Strengths & Weaknesses Recruiting and Screening At-Risk Strengths • Very smooth getting at-risk youth – returning protégés already fit the profile • There were no problems recruiting at-risk youth, because we were working with that population anyways. • The at-risk group is attracted to FNL anyway, we had a wonderful group. It was trying at times, but they were paired with mentors that matched them so well – I saw some great moments. Challenges • There were a lot more discipline problems such as lack of interest and lack of involvement. • There appears to be a stigma at this site about staying on campus after school hours • The screening process made finding these individuals more difficult because there are few students at the school who fit the criteria of a committed, good student who is not involved in conflicting extra-curricular activities
Reporting to Funders l l l Use the specified format Address the original grant initiatives as focal point Use language that links back to original proposal Highlight lessons learned Review sustainability
Sharing with Key Stakeholders l l l Be concise Use very basic statistics & graphics Make information accessible to broad audience Use exciting/interesting format Acknowledge contributions Highlight steps toward the future For Policy-makers: make specific recommendations l
Get the News Out l l l Newsletter Press Release Newspaper Articles Local Television Station Organized meetings (program staff; city council; school board; PTA)
Guide: Step 5 ü Audience-specific format
Activity: Ask the Wizard
Finally… You now know how to: Set up, code, enter, and clean data ü Translate data into findings Add context and interpretation to findings ü Disseminate evaluation findings ü ü
The End. (woo hoo!)
Скачать презентацию GOT DATA Step-by-Step Guide to Making Data Work
b774245af9cc9141c0cb3fb4c79e9490.ppt