全国高等学校英语教师 教育与发展系列研修班
DESCRIPTION
全国高等学校英语教师 教育与发展系列研修班. Statistics in action: A listening strategies project 25 July 2010 Victoria University of Wellington 中国外语教育研究中心 顾永琦 [email protected]. In this session…. Part 1: Feedback to your journals Part 2: Exercises 1 and 2 demo - PowerPoint PPT PresentationTRANSCRIPT
1
全国高等学校英语教师教育与发展系列研修班
Statistics in action: A listening strategies project
25 July 2010
Victoria University of Wellington中国外语教育研究中心
In this session…
Part 1: Feedback to your journals Part 2: Exercises 1 and 2 demo Part 3: Statistics in a listening
strategy project
Part 1
Your journals 收获 问题 建议
问题
运用实例,看什么研究问题用什么统计方法,生成什么样的图表,如何解读
一些关键术语还是模糊,如 nominal, ordinal, interval scales, p值,相关正负值
Comment : realistic expectations
建议
加强练习与实际操作 SPSS操练手册? 提供 references 提前把材料放在网上
Part 2
Exercises 1 and 2 demo
If you run Frequencies on the data, you get the descriptive statistics for all 44 cases, i.e., for both classes.
What you need is mean scores and SD for both classes.
You can get the descriptive statistics you want from the Explore command under Descriptive statistics.
Reporting your descriptive statistics
Class N 均值 标准差
Professor's Class 21 55.29 15.922
Graduate Student's Class 23 58.13 7.156
Interpreting your findings The mean score of the graduate student’s
class (Mean=58.13) was indeed higher than that of the professor’s class (Mean=55.29). In general, the graduate student’s class outperformed the professor’s class.
The professor’s class also had a much larger standard deviation than the student’s class, indicating that the professor’s class showed a wider spread of scores than the other class.
NB: you can also show the spread visually on a histogram.
Exercises 2
Correlation
Choose your options to include descriptive statistics
Report the descriptive statistics first
Report the correlation table
Interpreting your correlation findings
The findings showed that candidates’ abilities were positively correlated with the salary package offered to them. All four ability measures were found to be significantly correlated to the candidates’ annual salary.
The highest correlation existed between the candidates’ IQ scores and their salary (r=.844, p<.001). In other words, someone’s IQ score explains about 65% of the salary s/he earns. Communication score revealed the second highest correlation with salary (r=.813, p<.001)…
This is encouraging news for business school students, in that hard work, good marks, and an intelligent mind are indeed highly related to the salary they get after they graduate.
23
Part 3: Stats in action
3.1. Developing and validating the Strategy Battery for Young Learners of English (SBYLE)
3.2. Statistics in a listening strategy project
24
Strategy Battery for Young Learners of English (SBYLE)
Purpose: Profiling English language learner strategies used by upper primary school pupils in Singapore
1. Listening Strategy Questionnaire 2. Reading Strategy Questionnaire3. Writing Strategy Questionnaire
25
Designing the SBYLE The listening strategies questionnaire
as an example
26
Version numbers for the Strategy Battery for Young Learners of English
Version Purpose
1.0 Creation of questionnaire items
1.1 Validation by experts
1.2 Validation by school teachers
1.3 Validation by target sample (N=6)
1.4 •Piloting among large sample (N=293)•Ensuring internal consistency reliability
2.0 Final administration (N=3,618)
27
Version 1.0: Creation of questionnaire items
What was done Criteria1. Top-down: metacognitive,
cognitive, and social/affective
2. Bottom-up: Turn Phase I codings into questionnaire items.
3. List items in categories4. Find items in existing
literature, integrate them into what has been created, under existing or new categories.
1. Minimum 3, maximum 10 in each category, no single item variables. (strategies are latent variables: A strategy is not a single instance of behaviour)
2. Closed, 5-point Likert scale: Never to Always
3. Include a purpose, if possible, in each item. (strategies are purposeful)
28
Where do Questionnaire items come from? Phase I studies
Participants: 4 schools, all 6 grades in each school, one
high, one medium, and one low EL proficiency pupil from each grade (4x6x3=72)
Materials: Listening, reading, writing tasks Think-aloud protocols
Analysis: Coding and pattern exploration
29
Version 1.0 ExampleStrategy type Strategies Items
Metacognitive
Self-initiation
1. I look for opportunities to listen in English.
2. To improve my listening in English, I watch English TV programmes.
3. When I’m free, I find interesting things to listen to in English (for example, TV, radio, etc).
4. I try to find out how to improve my listening in English.
30
Validating the SBYLE
31
Validity checks Content validity
Consult experts, teachers, target group on 1) relevance 2) coverage 3) representativeness, and 4) exactness of wording
Construct validity Comparison with theory Factor analysis
Response validity Participating pretest: paired think-aloud
32
Reliability checksType of
reliabilityDecision
Internal consistency
reliability
yes
Test-retest reliability
no
33
Version 1.1: Validation by experts
What was done Guiding questions
Item-by-item examination and change
(What do we intend to cover?)
Conceptual coverage: •Do items under each category belong there? •Should there be other items?•Should there be other categories?
Wording: •Is wording for each item clear?•Is wording for each item simple enough to be understood by upper primary school pupils?
34
Changes made at this stage: Examples
Original Revised
When I listen to something, I try hard to understand every word before I can put these words together to understand the whole thing.
When I listen, I pay attention to every word.
When I listen, I try to form images in my mind for better understanding.
When I listen, I try to form pictures in my mind in order to understand better.
I try to relate what I hear to my own experiences in order to learn something from what I hear.
I try to see connections between what I hear and my experiences.
35
Version 1.2: Validation by teachers
What was done Guiding questions
Consultation with teachers
(intended meaning = perceived meaning?)
1. Is wording for each item clear? 2. Is wording for each item
simple enough to be understood by upper primary school pupils?
3. If the wording of an item is not clear enough, what alternative suggestions would you make?
36
Version 1.3: Validation by target sample (intended meaning = perceived meaning?)
What was done Guiding questions
1. Randomize items in each questionnaire
2. The formal version for small-scale piloting among 6 pupils item by item
3. Design answer sheets
1. Do pupils understand each item?
2. Are there items that are easily misunderstood?
3. Do pupils have problems with the answer sheet?
4. Are there any other snags?5. How much time is needed to
answer each questionnaire?
37
Version 1.4: Construct validity and internal consistency reliability
Formal piloting among 293 pupils (P4-P6)
Elicit student and teacher comments Revise, after factor analysis and item
analysis: delete, add, change, collapse categories, re-categorize items
Fine-tune survey administration procedures and detect problems (Guidelines for Questionnaire Administration)
38
Version 1.4: Reliability analysisVariable Item #
in pilot Q’naire
Item in original categoriesItem-Total correlation
Alpha if item deleted
Scale alpha
Inferencing 2
delete
When I don’t understand something, I use my knowledge about the English language to guess.
.371 .672 .696
6 When I don’t understand something, I use what I have already heard to guess.
.519 .623
22 When I don’t understand something, I make several guesses.
.395 .666
33 When I don’t understand something, I use my general knowledge to make a guess.
.538 .618
38
delete
When I don’t understand something, I use my knowledge about text structure to help me understand.
.218 .721
41 When I don’t understand something, I use my knowledge about the topic to guess.
.539 .618
39
Deleted item: #38Item in original categories Item-Total
correlation
Alpha if item deleted
When I don’t understand something, I use my knowledge about text structure to help me understand.
.218 .721
40
Deleted item #2
Item in original categories
Item-Total correlation
Alpha if item deleted
When I don’t understand something, I use my knowledge about the English language to guess.
.371 .672
Obtaining variables from individual items
COMPUTE LSELFINI=(L11+L12+L14 +L18)/4. COMPUTE LPLAN=(L3+L15+L23)/3. COMPUTE LMONITOR
=(L21+L32+L5+L9+L36)/5. COMPUTE LPERCEPT =(L8+L24 +L34+L37)/4. COMPUTE LINFER =(L1+L4+L17+L26
+L31+L27+L33)/7. COMPUTE LPREDICT
=(L6+L7+L13+L29+L35)/5. COMPUTE LUTILISE
=(L16+L19+L20+L22+L30)/5. COMPUTE
LSOCIAL=(L2+L10+L28+L25+L38)/5.
Running item analysis and reliability in SPSS
What we are doing:1. Checking if items that make up a variable
are consistent among themselves.2. Doing item-total correlation to see how
each item in the variable is correlated to the total variable score.
3. Determining which item should be removed from the variable (alpha is item deleted)
Example: the monitoring variable LMONITOR was made up of 5 items L21, L32, L5, L9, and L36.
Getting internal-consistency reliability
Locate all items in the variable
Specifying the statistics you want
Deciding if any item does not belong to the variable
48
Version 1.4: Problems and solutions
Administration problems (N=293)
Solutions
1.Pupils don’t remember ID numbers
2.Wrong understanding of “Grade”
3.Wrong answer sheets used
1.Ask schools to remind students in advance
2.Provide P4/P5/P6 choices3.Tell Teachers in advance
that different questionnaires require different answer sheets
49
Version 2.0: Final Administration
Number of Pupils
Percent
School 1 629 17.4School 2 403 11.1School 3 856 23.7School 4 762 21.1School 5 440 12.2School 6 528 14.6Total 3618 100.0
Statistics in the listening strategy project:
Deciding what stats to use Research question Statistics1. What is the listening strategy
use pattern among Singapore upper primary school pupils?
Descriptive statistics
2. How are listening strategies related to English language results?
Correlation, Regression (strength of associations between the two vars) orANOVA (locating differences between top, middle, & bottom groups)
3. Do strategies change from grade to grade?
ANOVA (differences among three grades)
Reporting descriptive statistics:
The table format
Interpretation and textual report
Listening strategies and English grades
Textual interpretation
Run your own item analysis and Chronbach’s alpha in SPSS Items 15, 17, 14, 25, 5, 8, 11, 34, 26 Items 1, 2, 3, 6, 18 Items 24, 29, 16, 19 Items 28, 27, 21 Items 30, 31, 32 Items 7, 10, 4, 33, 20 Items 23, 12, 22