Practical Connection Assignment
Kishan@1529
Week7presentation1.pptxReview:
ANOVA post-hoc tests
ANOVA
ANOVA results
The ANOVA indicated that there is a significant difference (p < .001), so post hoc tests are needed.
All combinations should be reviewed.
Run t-tests for a-b, a-c, a-d, b-c, b-d, c-d
The first three t-tests:
There is no difference between television (M = 52.89) and Internet (M =67.44), (t [16] = 1.29, p > .05).
There is a significant difference between television (M = 52.89) and print (M =29.67), (t [13] = 1.85, p < .05).
There is a significant difference between television (M = 52.89) and other (M =14.2), (t [12] = 2.82, p < .05).
The last three t-tests
There is a significant difference between Internet (M=67.44) and print (M= 29.67), (t [13] = 5.58, p < .001).
There is a significant difference between Internet (M=67.44) and other (M =14.2), (t [12] = 7.38, p < .001).
There is a significant difference between print (M= 29.67) and other (M =14.2), (t [9] = 5.10, p < .001).
Final Lesson:
Regression, Pearson Product Moment Correlation Coefficients, r
Reminder, please submit your article reviews prior to next Monday evening. The final will be released on Thursday. It is set with forced completion which means you will need to complete it in one sitting. It will need to be turned in by Monday at midnight.
7
Statistical Tests:
Regression
A regression is also used if you pilot test a survey. Suppose you develop a survey instrument or find a survey instrument to use, but it has not been proven reliable. In this case you would pilot test the survey and use a regression test to determine if the questions/categories are reliable. You would need a minimum of .80 on the regression test for reliability.
8
Regression
In statistical terms, we use “correlation” in reference to similarities between and among data sets.
A regression produces a Pearson’s Product Moment Correlation r.
The regression is best used when one can assume that the distribution of scores in both groups forms a single mode distribution.
In the Spatz appendices, use Table A: Critical values for Pearson product-moment correlation coefficients, r (to determine significance).
We are looking for significance at the .05 level (p < .05) to determine if there is a correlation.
9
More about regression:
Correlation coefficients range from -1.00 to 1.00.
A correlation coefficient of .00 means there is no cause and effect relationship, but a correlation coefficient of 1.00 means there may be a cause and effect relationship.
Note the degrees of freedom is N-2.
In a bivariate distribution both variables have the same number of scores.
Positive correlation: the larger the building, the more the cost of upkeep (both increase).
Negative correlation: the more I eat the less I weigh (one increases while the other decreases).
Scatter plots are on page 101 in your textbook.
10
Dissertation example:
A regression test was used to measure the relationship between teachers’ job satisfaction and student achievement scores as reflected in their schools’ academic index. According to the regression test, there is a significant relationship between teacher job satisfaction and student test scores, (r [48] = .60, p < .01) (Spatz, 2008). Table 4.1 shows the results of the regression test.
| Regression Statistics | |
| Multiple R | 0.604572038 |
| R Square | 0.365507349 |
| Adjusted R Square | 0.352288752 |
| Standard Error | 0.594867874 |
| Observations | 50 |
| F [1,48]= 27.65, p < .01 |
Page 102 of your text provides effect size explanation. In this case, an r of .60 would fall in the large effect size category. The coefficient of determination measures common variance (see page 102 in your text). To determine common variance it would be.60 squared. What is that value? .36 Calculating the coefficient of determination and effect size is not necessary for your final. However, for your dissertation or publishing you may be asked t calculate this.
Even though we have a high correlation, we can’t say there is cause and effect. It may be the result of cause and effect, more satisfied teachers make scores higher or schools with higher scores make teachers more satisfied.
11
Regression Practice
A researcher gave 9 participants two different tests that measure self-esteem to see if there was a relationship between the tests. Are they significantly similar?
Steps:
Enter data
Click the “Data” tab.
Select “Data Analysis,” on the right
Select “Regression”
Click “Ok.”
12
Next…
6. Highlight the 1st data column for “Input Y Range.”
7. Highlight the 2nd data column for the “Input X Range.”
8. Click “Labels,” if you labeled columns.
9. Select a cell for output range.
10. For a graph, click “Line Fit Plots.”
11. Click “Ok.”
13
The Results
According to page 24 of the dissertation handbook, report the results as: (r = .84, F [1,7] = 16.96, p < .01).
The results can also be reported as: (r [7] = .84, p < .01).
Either way, this means that there is a significant relationship between subjects self-esteem scores.
The handbook shares two ways to report the regression, so I’m sharing both with you. The second way, without the F statistic is the most commonly reported in our dissertations. The degrees of freedom is n-2.
14
See in the Spatz Table:
(r [7] = .84, p < .01)
15
Same data with the Pearson’s Product Moment r
(r [7] = .84, p < .01)
