Stats - SPSS

Question 1

When you have treatment readings from two sample groups and want to compare them…

• Test type
• SPSS menu option
• 4 key steps to reading the results
• Format to present the result

Answer 1

• Independent groups t-test
• Analyze –> Compare Means –> Independent Samples t-test
• 4 key steps
  
  1. check the coding of the IV groups (entered in correct order)
  2. Equality of variance? (Levene’s test: if the F-statistic is significant (p < 0.05), then you have to use
  3. Check the direction of the difference (is group A higher/lower than group B)
  4. Is the difference between the groups significant?
• t(n) = , p < 0.###
  
  • where n = degrees of freedom

Question 2

When you have treatment readings from one sample groups and want to compare it to a known mean for another sample.

• Test type
• SPSS menu option
• things to remember
• Format to present the result

Answer 2

• 1-sample T test
• Analyze –> Compare means –> One sample t-test
• order should be DV - IV
• t(n) = , p < 0.###
  
  • where n = degrees of freedom

Question 3

When you have treatment readings from participants (pre- and post-test) and want to compare them…

• Test type
• SPSS menu option
• 4 key steps to reading the results
• Format to present the result

Answer 3

• Paired Samples t-test
• Analyze –> Compare Means –> Paired Samples t-test
• 4 key steps
  
  1. Add DV readings to variables 1 and 2 - only need to run 1 pair
  2. Equality of variance? (Levene’s test: if the F-statistic is significant (p
  3. Check the direction of the difference (is group A higher/lower than group B)
  4. Is the difference between the groups significant?
• t(n) = , p
  
  • where n = degrees of freedom

Question 4

1. Menu option
2. 3 key things
3. Degrees of freedom - how calculated?
4. Report as?

Answer 4

1. Analyze –> Descriptive Statistics –> Crosstabs
2. 3 key things
   
   1. doesn’t matter which way you tabulate (cols vs rows)
   2. select ‘Statistics’ and check ‘chi square’ option
   3. Go to ‘cells’ and add counts (observed, expected) + percentages
3. (r-1)(c-1)
4. chi^2(df) = #.###, p = o.###

Question 5

What are the 3 steps to running a chi^2 test where the data are already summarised as totals?

Answer 5

1. Code IVs in separated columns (remember to add in the variable view)
2. Add the frequencies in a 3rd column
3. Weight cases by frequency
   
   • (Data –> Weight Caes –> Weight by the Frequency variable)

Question 6

• SPSS menu option
• 2 key pre-flight checks
• What goes where in the dialogue box?
• Format to present the result

Answer 6

• Analyze –> Compare Means –> One-way ANOVA
• 2 key pre-flight checks
  
  1. Confirm your IV and DV
  2. Check your IV has appropriate labels
• IV goes into FACTOR, DV goes into DEPENDENT LIST
• F(df_bg, df_wg) = #.###, p =0.###
  
  • df_bg = between-groups d.o.f;
  • df_wg = within-groups d.o.f. (error d.o.f.)

Question 7

• SPSS menu option (including plots)
• 3 key steps to reading the results
• Format to present the result

Answer 7

• Analyze –> General Linear Model –> Univariate
  
  • Plots –> add one IV as horizontal, another as separate lines
• 3 steps:
  
  • Corrected Model = significance of the general model
  • Read each IV’s F-stat for MAIN EFFECTS
  • Read IV1*IV2 F-stat for INTERACTION
• F(df_bg, df_wg) = #.###, p =0.###
  • df_bg = between-groups d.o.f;
  • df_wg = within-groups d.o.f. (error d.o.f.)

Question 8

• SPSS menu option
• 5 key steps to reading the results
• Format to present the result

Answer 8

• Analyze –> General Linear Model –> Repeated Measures
• 5 key steps
  
  1. Assign factor name (this is just a description) and number of levels (number of repeated measures)
  2. Assign within-subjects factors in a suitable order
  3. If you have a between-subjects factor (e.g. Gender) don’t forget to add it
  4. Add plots
     
     • ​​Plots –> within-subjects factor on horizontal, between-subjects as separate lines
  5. ​Add a check for sphericity (Mauchly’s)
• F(df_bg, df_wg) = #.###, p =0.###
  • df_bg = between-groups d.o.f.
  • df_wg = within-groups d.o.f. (error d.o.f.)

Question 9

• When do you run post-hoc tests?
• How do you know what kind to use?
• Method in SPSS?

Answer 9

• When you don’t have an a-priori hypothesis but want to compare different treatment effects of the IV.
• You need to test for homogeneity of variance across the treatmeng groups, i..e a Levene’s Test.
  
  • If not significant, you can use a Tukey Test
  • If significant, use Games-Howell
• Method:
  
  • Options –> Homogeneity ofVariance Test
  • Post-Hoc –> Tukey (under EVA) and Games-Howell (under EVNA)

Question 10

• SPSS menu option
• 3 key checks to perform
• What is your p-value threshold?

Answer 10

• Options –> select Factor, check COMPARE MAIN EFFECTS, select BONFERRONI CORRECTION
• 3 key checks:
  
  • Look at Pairwise Comparisons table
  • Remember to pay careful attention to the direction of the mean difference
  • Check significance of each pair (note repetition)
• Bonferroni adjustment here is implict. So this means in this case use p

Question 11

• What test do you use?
  
  • Name, statistic, decision you need to make
• When would you use it?

Answer 11

• Levene’s test
  
  • F statistic. If significant (p < 0.05) then you cannot assume homogeneity of variance.
• independent samples T-test; One-way ANOVA post-hoc analyses.

Question 12

• What test should you run to test whether your data follows a normal distribution?
• How do you run it?
• Any caveats?

Answer 12

• Kolmogorov-Smirnoff test
• Descriptive Statistics –> Plots –> Normality plots with tests
• Can be misleading if:
  
  • running against a large sample (will detect signficance in small or unimportant effects)
  • using a small sample, as will lack power to detect violations

Question 13

• What is your trend analysis doing?
• How many models can it test for?
• SPSS menu option?
• 3 key steps for analysing trends.

Answer 13

• This is a form of contrast analysis. It is attempting to fit a trend model to the data and giving an indication if that model is significant
• One less than number of within-subjects treatments
• Contrasts –> polynomial
• Steps:
  
  1. Check ‘Tests of within-subjects Contrasts’ table
  2. Calculate your p-value threshold (i.e. do a Bonferroni correction: 0.05 / (num. of trends)
  3. See how many trends have significance (where p-value is less than your p-value threshold)

Question 14

• How do you get descriptive statistics?

Answer 14

• Analyze –> Descriptive Statistics –> Explore
• Take both stats and plots
• To get the test of normality of data, go to Plots –> check box for normality tests

Question 15

[]

Question 16

• When do you run Contrasts?
• Method?
• Menu options

Answer 16

• If you have an a-priori hypothesis
• Need to run contrasts with a Bonferroni correction
  *

Question 18

• What is skewness?
• How can you look for it (data + visuals)

Answer 18

• Sample data does not sit ‘evenly’ about a mean. Results tend to cluster to the left or right of your observed data range.
• Positive number = +ve skew, i.e. results tend to be grouped to the left hand side.
• Negative number = -ve skew, i.e. results tend to be grouped to the right hand side
• Analyze –> Descriptive statistics. The resulting output has skewness measures (in s.d.s). You can also create histograms.
  *

Question 21

[]

Question 22

[