《社会科学研究方法》课程教学资源：SPSS Software_Statistics-Spss Guide For Dummies.pdf_大学文库

ii TABLE OF CONTENTS PAGE Helpful Hints for All Tests................................................................................................1 Tests for Numeric Data 1. Z-Scores.................................................................................................................1 2. Helpful Hints for All T-Tests.................................................................................2 3. One Group T-Tests.................................................................................................2 4. Independent Groups T-Test ...................................................................................3 5. Repeated Measures (Correlated Groups or Paired Samples) T-Test .....................3 6. Independent Groups ANOVA................................................................................4 7. Repeated Measures (Correlated Groups or Paired Samples) ANOVA .................4 8. Correlation Coefficient ..........................................................................................4 9. Linear Regression ..................................................................................................5 Tests for Ordinal Data 1. Helpful Hints for All Ordinal Tests .......................................................................7 2. Kruskal – Wallis H.................................................................................................7 3. Friedman’s .............................................................................................................7 4. Spearman’s.............................................................................................................7 Tests for Nominal Data 1. Helpful Hints for All Nominal Tests.....................................................................8 2. Chi-Square Goodness-of-Fit ..................................................................................8 3. Chi-Square Independence ......................................................................................8 4. Cochran’s Q ...........................................................................................................8 5. Phi or Cramer’s V (Correlations for Nominal Data) .............................................9

SPSS Guide to Data Analysis Page 1 of8 For All Tests Remember that the Significance (or Asymp.Sig.in some cases)needs to be less than 0.05 to be significant. The Independent Variable is always the variable that you are predicting something about (i.e.what your Ha predicts differences between,as long as your Ha is correct).The Dependent Variable is what you are measuring in order to tell if the groups (or conditions for repeated measures tests)are different.For correlations and for Chi-Square,it does not matter which one is the Independent or Dependent variable. Ha always predicts a difference (for correlations,it predicts that r is different from zero,but another way of saying this is that there is a significant correlation)and Ho always predicts no difference.If your Ha was directional,and you find that it was predicted in the wrong direction(i.e.you predicted A was greater than B and it turns out that B is significantly greater than A)you should still accept H,even though H predicts no difference,and you found a difference in the opposite direction. If there is a WARNING box on your Output File,it is usually because you used the wrong test,or the wrong variables.Go back and double check. Tests For Numeric Data Z-Scores (Compared to Data) Analyze Descriptive Statistics>Descriptives Click over the variable you would like zscores for Click on the box that says Save Standardized Values as Variables.This is located right below the box that displays all of the variables. If means and standard deviations are needed,click on Options and click on the boxes that will give you the means and standard deviations. .The zscores will not be on the Output File!!! They are saved as variables on the Data File.They should be saved in the variable that is to the far right of the data screen.Normally it is called z,and then the name of the variable (e.g.ZSLEEP) Compare the zscores to the critical value to determine which zscores are significant.Remember,if your hypothesis is directional(i.e.one-tailed),the critical value is or-1.645.If your hypothesis is non-directional (i.e.two- tailed),the critical value is or-1.96

SPSS Guide to Data Analysis Page 1 of 8 For All Tests · Remember that the Significance (or Asymp. Sig. in some cases) needs to be less than 0.05 to be significant. · The Independent Variable is always the variable that you are predicting something about (i.e. what your Ha predicts differences between, as long as your Ha is correct). The Dependent Variable is what you are measuring in order to tell if the groups (or conditions for repeated measures tests) are different. For correlations and for Chi-Square, it does not matter which one is the Independent or Dependent variable. · Ha always predicts a difference (for correlations, it predicts that r is different from zero, but another way of saying this is that there is a significant correlation) and Ho always predicts no difference. If your Ha was directional, and you find that it was predicted in the wrong direction (i.e. you predicted A was greater than B and it turns out that B is significantly greater than A) you should still accept Ho, even though Ho predicts no difference, and you found a difference in the opposite direction. · If there is a WARNING box on your Output File, it is usually because you used the wrong test, or the wrong variables. Go back and double check. Tests For Numeric Data Z-Scores (Compared to Data ) Analyze ‡ Descriptive Statistics ‡ Descriptives · Click over the variable you would like z-scores for · Click on the box that says Save Standardi zed Values as Variables. This is located right below the box that displays all of the variables. · If means and standard deviations are needed, click on Options and click on the boxes that will give you the means and standard deviations. · The z-scores will not be on the Output File!!! · They are saved as variables on the Data File. They should be saved in the variable that is to the far right of the data screen. Normally it is called z, and then the name of the variable (e.g. ZSLEEP) · Compare the z-scores to the critical value to determine which z-scores are significant. Remember, if your hypothesis is directiona l (i.e. one-tailed), the critical value is + or – 1.645. If your hypothesis is non-directional (i.e. twotailed), the critical value is + or – 1.96

SPSS Guide to Data Analysis Page 3 of 9 the "Percent Correct on a Test"variable were entered as 0.80,0.75,etc.,then the vest value should be entered as 0.78.If the data were entered as 90,75, etc.,then the test value should be entered as 78.In order to know how the data were entered,click on the Data File screen and look at the data for the dependent variable For Independent Groups T-Tests Analyze>Compare Means Independent-Samples T Test The Dependent Variable goes in the Test Variable box and the Independent variable goes in the Grouping Variable box. Click on Define Groups and define them.In order to know how to define them, click on Utilities>Variables.Click on the independent variable you are defining and see what numbers are under the value labels (i.e.usually it's either 0 and 1 or I and 2).If there are more than two numbers in the value labels then you cannot do a t-test unless you are using a specified cut-point(i.e.if there are four groups:1 =old women,2 young women,3=old men,4=young men,and you simply wanted to look at the differences between men and women,you could set a cut point at 2).If there are no numbers,you should be using a specified cut- point.If you have more than two numbers in the value label,or if you have no numbers in the value labels,and no cut-point has been specified on the final exam you are doing the wrong kind of test!!!! On the Output File:Remember,this is a t-test,so ignore the F value and the first significance value (Levene's Test).Also,ignore the equal variances not assumed row. Before accepting Ha,be sure to look at the means!!!If I predict that boys had a higher average correct on a test than girls and my t-test value is significant I may say,yes,boys got more correct than girls.However,this t-test could be significant because girls got significantly more correct than boys!!(Therefore,Ha was predicted in the wrong direction!)In order to know which group got significantly more correct than the other,I need to look at the means and see which one is bigger!！ For Repeated Measures (a.k.a Correlated Groups,Paired Samples)T-Test Analyze>Compare Means>Paired-Samples T Test ● Click on one variable and then click on a second variable and then click on the arrow that moves the pair of variables into the Paired Variables box.In order to make things easier on yourself,click your variables in,in order ofyour hypotheses.For example,ifyou are predicting that A is greater than B,click on A first.This way,you should expect that your t value will be positive.If the test is significant,but your t value is negative,it means that B was significantly greater than A!!(So Ha was predicted in the wrong direction and you should accept Ho). Ifyou are predicting that A is less than B,then click on A first.This way,you

SPSS Guide to Data Analysis Page 3 of 9 the “Percent Correct on a Test” variable were entered as 0.80, 0.75, etc., then the vest value should be entered as 0.78. If the data were entered as 90, 75, etc., then the test value should be entered as 78. In order to know how the data were entered, click on the Data File screen and look at the data for the dependent variable. For Independent Groups T-Tests Analyze ‡ Compare Means ‡ Independent-Samples T Test · The Dependent Variable goes in the Test Variable box and the Independent variable goes in the Grouping Variable box. · Click on Define Groups and define them. In order to know how to define them, click on Utilities ‡ Variables. Click on the independent variable you are defining and see what numbers are under the value labels (i.e. usually it’s either 0 and 1 or 1 and 2). If there are more than two numbers in the value labels then you cannot do a t-test unless you are using a specified cut-point (i.e. if there are four groups: 1 = old women, 2 = young women, 3 = old men, 4 = young men, and you simply wanted to look at the differences between men and women, you could set a cut point at 2). If there are no numbers, you should be using a specified cutpoint. If you have more than two numbers in the value labels, or if you have no numbers in the value labels, and no cut-point has been specified on the final exam you are doing the wrong kind of test!!!! · On the Output File: Remember, this is a t-test, so ignore the F value and the first significance value (Levene ’s Test). Also, ignore the equal variances not assumed row. · Before accepting Ha, be sure to look at the means!!! If I predict that boys had a higher average correct on a test than girls and my t-test value is significant I may say, yes, boys got more correct than girls. However, this t-test could be significant because girls got significantly more correct than boys!! (Therefore, Ha was predicted in the wrong direction!) In order to know which group got significantly more correct than the other, I need to look at the means and see which one is bigger!! For Repeated Measures (a.k.a Correlated Groups, Paired Samples) T- Test Analyze ‡ Compare Means ‡ Paired-Samples T Test · Click on one variable and then click on a second variable and then click on the arrow that moves the pair of variables into the Paired Variables box. [In order to make things easier on yourself, click your variables in, in order of your hypotheses. For example, if you are predicting that A is greater than B, click on A first. This way, you should expect that your t value will be positive. If the test is significant, but your t value is negative, it means that B was significantly greater than A!! (So Ha was predicted in the wrong direction and you should accept Ho). If you are predicting that A is less than B, then click on A first. This way, you

SPSS Guide to Data Analysis Page 4 of 9 should expect that the t value will be negative.If the test is significant but the t value is positive,you know that it means that B was significantly less than A (so Ha was predicted in the wrong direction,and you should accept Ho). For Independent Groups ANOVA Analyze→General Linear Model→One-Vay ANOVA Put the Dependent variable into the Dependent List box.Put the Independent variable into the Factor box. ● Click on the Post Hoc box.Click on the Tukey box.Click Continue If means and standard deviations are needed,click on the Options box.Then click on Descriptive. For Repeated Measures (a.k.a.Correlated Groups,Paired Samples)ANOVA Analyze General Linear Model>Repeated Measures Type in the within-subject factor name in the Within-Subject Factor Name box. This cannot be the name of a pre-existing variable.You will have to make one up. Type in the number of levels in the Number of levels box.How do youknow how many levels there are?If my within-subject factor was "Tests"and I have a variable called "Test 1"a variable called"Test 2"and a variable called"Test 3," then I would have 3 levels.In other words,the number of levels equals the number of variables(that you are examining)that correspond to the within- subject factor. 。Click on Define Put the variables you want to test into the Variables box.Preferably,put them in the right order(if there is an order to them).This will keep you from getting confused.For example,I should put in my "Test 1"variable in first,my "Test 2" variable in second,etc. For post hoc tests,click on Options,highlight the variable,move it into Display Means For box,click on Compare Main Effects,change Confidence Interval Adjustment to Bonferonni (the closest we can get to Tukey's Test).You may also want to click on Estimates of Effect Size for eta Remember to look at the Tests of Within-Subjects Effects box for your ANOVA results. Correlation Coefficient(r) Analyze→Correlate→Bivariate Make sure that the Pearson box (and only the Pearson box)has a check in it. Put the variables in the Variables box.Their order is not important

SPSS Guide to Data Analysis Page 4 of 9 should expect that the t value will be negative. If the test is significant but the t value is positive, you know that it means that B was significantly less than A (so Ha was predicted in the wrong direction, and you should accept Ho).] For Independent Groups ANOVA Analyze ‡ General Linear Model ‡ One-Way ANOVA · Put the Dependent variable into the Dependent List box. Put the Independent variable into the Factor box. · Click on the Post Hoc box. Click on the Tukey box. Click Continue . · If means and standard deviations are needed, click on the Options box. Then click on Descriptive . For Repeated Measures (a.k.a. Correlated Groups, Paired Samples) ANOVA Analyze ‡ General Linear Model ‡ Repeated Measures · Type in the within-subject factor name in the Within-Subject Factor Name box. This cannot be the name of a pre-existing variable. You will have to make one up. · Type in the number of levels in the Number of levels box. How do you know how many levels there are? If my within-subject factor was “Tests” and I have a variable called “Test 1” a variable called “Test 2” and a variable called “Test 3,” then I would have 3 levels. In other words, the number of levels equals the number of variables (that you are examining) that correspond to the withinsubject factor. · Click on Define . · Put the variables you want to test into the Variables box. Preferably, put them in the right order (if there is an order to them). This will keep you from getting confused. For example, I should put in my “Test 1” variable in first, my “Test 2” variable in second, etc. · For post hoc tests, click on Options, highlight the variable, move it into Display Means For box, click on Compare Main Effects, change Confidence Interval Adjustment to Bonferonni (the closest we can get to Tukey’s Test). You may also want to click on Estimates of Effect Size for eta2 . · Remember to look at the Tests of Within-Subjects Effects box for your ANOVA results. Correlation Coefficient (r) Analyze ‡ Correlate ‡ Bivariate · Make sure that the Pearson box (and only the Pearson box) has a check in it. · Put the variables in the Variables box. Their order is not important

SPSS Guide to Data Analysis Page 5 of 9 Select your tailed significance (one or two-tailed)depending on your hypotheses Remember directional hypotheses are one-tailed and non-directional hypotheses are two-tailed. If means and standard deviations are needed,click on Options and then click on Means and Standard Deviations Linear Regression Graphs→Scatter→Simple This will let you know if there is a linear relationship or not. Click on Define. The Dependent variable (criterion)should be put in the Y-axis box and the Independent variable(predictor)should be put in the X-axis box and hit OK .In your Output:Double click on the scatterplot.Go to Chart>Options.Click on Total in the Fit Line box,then OK. Make sure it is more-or-less linear.The next step is to check for normality Graphs→Q-Q Put both variables into the Variable box.Hit OK. .Look at the Normal Q-Q Plot of ....not the Detrended Normal Q-Q of... box.If the points are a "smiley face"they are negatively-skewed.This means you will have to raise them to a power greater than one.If the points make a "frown face"then they are positively-skewed.This means you will have to raise them to a power less than one (but greater than zero).To do this,go to the DATA screen then: Transform→Compute Give the new variable a name in the Target Variable box.Since you will be doing many of these (because it is a guess and check)it may be easiest to name it the old variable and then the power you raised it to.For example,SLEEP.2 if I raised it to the.2 power,or SLEEP3 if I raised it to the third power. Click the old variable(that you want to change)into the Numeric Expressions box.Type in the exponent function (**)and then the power you want to raise it to.Hit OK. Redo the Q-Q plot with the NEW variable (i.e.SLEEP.2,and not SLEEP) Repeat until you have the best fit data.The variable that you created with the best-fit data (i.e.SLEEP.2)will be the variable that you will use for the REST OF THE REGRESSION(no more SLEEP). The next step is to remove Outliers.To do so,run a regression: Analyze→Regression→Linear

SPSS Guide to Data Analysis Page 5 of 9 · Select your tailed significance (one or two-tailed) depending on your hypotheses. Remember directional hypotheses are one-tailed and non-directional hypotheses are two-tailed. · If means and standard deviations are needed, click on Options and then click on Means and Standard Deviations. Linear Regression Graphs ‡ Scatter ‡ Simple · This will let you know if there is a linear relationship or not. · Click on Define . · The Dependent variable (criterion) should be put in the Y-axis box and the Independent variable (predictor) should be put in the X-axis box and hit OK. · In your Output: Double click on the scatterplot. Go to Chart ‡ Options. Click on Total in the Fit Line box, then OK. · Make sure it is more-or-less linear. The next step is to check for normality. Graphs ‡ Q-Q · Put both variables into the Variable box. Hit OK. · Look at the Normal Q-Q Plot of …., not the Detrended Normal Q-Q of … box. If the points are a “smiley face” they are negatively-skewed. This means you will have to raise them to a power greater than one. If the points make a “frown face” then they are positively-skewed. This means you will have to raise them to a power less than one (but greater than zero). To do this, go to the DATA screen then: Transform ‡ Compute · Give the new variable a name in the Target Variable box. Since you will be doing many of these (because it is a guess and check) it may be easiest to name it the old variable and then the power you raised it to. For example, SLEEP.2 if I raised it to the .2 power, or SLEEP3 if I raised it to the third power. · Click the old variable (that you want to change) into the Numeric Expressions box. Type in the exponent function (**), and then the power you want to raise it to. Hit OK. · Redo the Q-Q plot with the NEW variable (i.e. SLEEP.2, and not SLEEP). · Repeat until you have the best fit data. The variable that you created with the best-fit data (i.e. SLEEP.2) will be the variable that you will use for the REST OF THE REGRESSION (no more SLEEP). · The next step is to remove Outliers. To do so, run a regression: Analyze ‡ Regression ‡ Linear