FOR SIGNIFICANCE LOOK AT PAIRED SAMPLES CORRELATION

PAIRED SAMPLES EFFECT SIZE = EFFECTIVNESS OF THERAPY

(LOOK AT POINT ESTMATE)

POINT ESTIMATE = CALCULATED EFFECT SIZE

 Insignificant: Hedges g < 0.20

 Moderate: 0.20 ≤ Hedges g < 0.50
 Strong: Hedges g ≥ 0.50

1.Independent t-test:

- Purpose: Compare means between two independent groups.

- SPSS Steps: Analyze > Compare Means > Independent Samples T-test.
- Report:
- df (degrees of freedom)
- t-value (test statistic)
- p-value (significance level; p < .05 is significant)
- Effect Size (Cohen’s d): d = (Mean1 - Mean2) / SD_pooled.

2. Paired t-test:

- Purpose: Compare means within the same group at two time points (e.g., pre- and post-test).
- SPSS Steps: Analyze > Compare Means > Paired Samples T-test.
- Report:
- df, t-value, p-value.
- Effect Size: Use Cohen’s d for paired samples: d = (Mean difference) / SD difference.

3. Correlation (Pearson’s r):

- Purpose: Measure the strength and direction of the relationship between two continuous
variables.
- SPSS Steps: Analyze > Correlate > Bivariate.
- Report:
- r-value: (between -1 and +1)
- p-value: Significance (p < .05 = significant)
- Effect Size: The r-value itself serves as the effect size:
- Small (r = .1), Medium (r = .3), Large (r = .5).
4. Linear Regression:

- Purpose: Predict the value of a dependent variable based on one or more independent
variables.
- SPSS Steps: Analyze > Regression > Linear.
- Check Descriptives: Mean, standard deviation, etc.
-Residual stats: CHECK MIN/MAX. THE
- R² (R2/R2 adjusted!): Indicates the proportion of variance in the DV explained by the IV(s).
Report R² to 2 decimal places.
- Durbin-Watson: Test for autocorrelation (0-4 range; close to 2 is ideal IF IT IS –
ASSUMPTION OF RESIDUALS ISN’T VIOLATED).
- ANOVA Table: Check significance of F-statistic to see if the model fits better than baseline (p
< .05).
- df1 = Number of predictors; df2 = N - df1 - 1.
- CHECK COEFFICIENTS TABLE FOR SIGNIFICANCE OF VARIABLES, RELATIONSHIPS:
- Unstandardized B: Represents change in DV for each unit change in IV.
- Standardized Beta (β): For comparing relative strength of predictors.
- Significance (p-value): p < .05 means predictor is significant.

MULTICOLLIN (ASSUMPTION)

- VIF: Checks for multicollinearity (closer to 1 is better; VIF > 10 is problematic).

- Tolerance: Closer to 1 is best (tolerance < .2 suggests multicollinearity, < .01 is

terrible).

-REGULAR CORRELATION (IF IT’S > 0.8, IT’S A PROBLEM)

IMPORTANT

CHECK SIGNIFICANT F CHANGE FOR 2 MODELS. REPORT IT, DF, F-VALUE

FIRST/SECOND MODEL AND WHETHER THEY ARE SIGNIFICANTLY BETTER THAN THE
BASELINE MODEL (SIG. F CHANGE)

LOOK AT SIGNIFICANT PREDICTOR IN COEFFICIENT TABLE. REPORT SIG, T,

UNSTARDASIED B AND STANDARDISED B.
THE SIGNIFICANT PREDICTORS ARE FOLLOWING, UNSTANDARDISED B (REPORT), IF
>1 SIGNIFICANT PREDICTOR – REPORT THE ONE WITH HIGHER UNSTANDARDISED B
(ONE IS MORE IMPORTANT)

HIGHER STANDARDISED B -> STRONGER IS CHARACTERISTIC

5. Logistic Regression:

- Purpose: Predict a binary outcome (e.g., success/failure).

- SPSS Steps: Analyze > Regression > Binary Logistic.
- Classification Table: Accuracy of predictions (% correctly predicted).
- Omnibus Test: Shows whether your model is better than baseline (p < .05).

GOOD MODEL FIT

- Hosmer-Lemeshow Test: Good fit if p (Sig) > .05. Can also report Nagelkerke R²
(pseudo R²). + CAN REPORT ALL 3 FROM MODEL SUMMARY
-CAN EVALUATE FROM HOSMER + OMNIBUS, IF ALL GOOD – REPORT CLASIFICATION
TABLE AS WELL (IT CORRECTLY CLASSIFIES (WHATEVER NUMBER IS IN THE VERY
BOTTOM)) + IT’S BETTER BECAUSE IT REDUCES (WHATEVER THE 2 ND LINE SAYS)
COMPARING TO BASELINE MODEL

DIRECTION OF RELETAIONSHIP
REPORT B, EXP (B), WALD, CI

EXAMPLE: ODDS INCREASE (EXP (B) IS 1.373 = 37.3%)

- Model Summary: Report -2 Log Likelihood, Cox & Snell R², and Nagelkerke R² (pseudo R²).
- Exp(B): Report odds ratios for each predictor (Exp(B)).
Key Points About Durbin-Watson:

 Range: The Durbin-Watson statistic ranges from 0 to 4.

o A value near 2 indicates no autocorrelation (i.e., the residuals are uncorrelated).
o A value closer to 0 indicates positive autocorrelation, meaning that adjacent
residuals are correlated positively (e.g., an error at one point tends to be followed
by a similar error at the next point).
o A value closer to 4 indicates negative autocorrelation, where adjacent
residuals are correlated negatively (e.g., an error at one point is likely to be
followed by an error in the opposite direction).

Interpreting Durbin-Watson:
 DW ≈ 2: Residuals are independent; no autocorrelation is detected.
 DW < 2: Indicates positive autocorrelation in the residuals.
 DW > 2: Indicates negative autocorrelation in the residuals.

RESIDUALS

Typical Thresholds
 Standardized residuals typically follow a normal distribution:
o Values close to 0 indicate that the observation is close to the predicted value.
o Values greater than +2 or less than -2 may indicate outliers, and values greater
than +3 or less than -3 are often considered highly unusual.

Conclusion
 In summary, minimum and maximum standardized residuals provide insights into the
behavior of individual data points in relation to your regression model:
o Minimum: Highlights observations that are significantly lower than predicted.
o Maximum: Highlights observations that are significantly higher than predicted.
 ASUMPTIONS

Linear Regression
Key Assumptions:
 Linearity: The relationship between the independent and dependent variables is linear.
 Independence: The residuals (errors) are independent (no autocorrelation). Checked
using the Durbin-Watson test.
 Homoscedasticity: The residuals have constant variance across all levels of the
independent variables (checked with a scatterplot of residuals vs. predicted values).
 Normality of Residuals: The residuals should be approximately normally distributed.
Checked using histograms, normal probability plots, or the Shapiro-Wilk test.
 No Multicollinearity: The independent variables are not too highly correlated. Checked
using VIF (Variance Inflation Factor).

Steps for Running Linear Regression:

1. Open Data:
o Go to File > Open and load your dataset.
2. Run the Regression:
o Go to Analyze > Regression > Linear.
o Move your dependent variable (the outcome you want to predict) to the
Dependent box.
o Move your independent variables (the predictors) to the Independent(s) box.
o Click OK.

Checking Key Assumptions:

1. Linearity:
o After running the regression, go to Graphs > Legacy Dialogs > Scatter/Dot.
o Choose Simple Scatterplot.
o Select your dependent variable for the Y-axis and independent variable for the
X-axis.
o Look for a straight line pattern in the scatterplot. A non-linear pattern would
indicate a violation of the linearity assumption.
2. Independence (Durbin-Watson Test):
o In the regression dialog (Analyze > Regression > Linear), click Statistics.
o Check the box for Durbin-Watson.
o A value close to 2 indicates independence of errors (no autocorrelation).
3. Homoscedasticity:
o After running the regression, go to Graphs > Legacy Dialogs > Scatter/Dot.
o Choose Simple Scatterplot.
o Use standardized predicted values for the X-axis and standardized residuals
for the Y-axis.
o Look for a random pattern. If the spread of residuals fans out or narrows,
homoscedasticity is violated.
 4. Normality of Residuals:
o After running the regression, go to Analyze > Descriptive Statistics > Explore.
o Move the Standardized Residual to the Dependent List.
o Go to Plots and check Normality plots with tests (you can also request a
histogram here).
o Review the P-P plot: residuals should closely follow the diagonal line for
normality.
o A Shapiro-Wilk test result should not be significant for normality.
5. No Multicollinearity:
o In the regression dialog (Analyze > Regression > Linear), click Statistics.
o Check the box for Collinearity diagnostics.
o SPSS will generate VIF (Variance Inflation Factor) values. VIF should be less
than 5 (ideally below 3). Higher values indicate multicollinearity.

Logistic Regression
Key Assumptions:
 Linearity of Logits: The logit transformation of the dependent variable should have a
linear relationship with the independent variables.
 Independence: The observations are independent.
 No Multicollinearity: The independent variables should not be highly correlated.
Checked using VIF.
 Adequate Sample Size: Logistic regression requires a larger sample size than linear
regression.

Steps for Running Logistic Regression:

1. Open Data:
o Go to File > Open and load your dataset.
2. Run the Logistic Regression:
o Go to Analyze > Regression > Binary Logistic.
o Move your dependent (binary outcome) variable to the Dependent box.
o Move your independent variables (predictors) to the Covariates box.
o Click OK.

Checking Key Assumptions:

1. Heteroscedasticity:
o SPSS does not automatically provide a test for this assumption.
o To check this, you can create interaction terms between the independent
variables and their logs, and include them in a model to test whether they are
significant.
o Alternatively, ensure your continuous predictors have a linear relationship with
the log-odds by plotting logit transformations manually.
2. Independence:
 o This is generally satisfied through proper study design (e.g., random sampling).
Ensure your data collection process respects independence.

3. No Multicollinearity:
o In the logistic regression dialog (Analyze > Regression > Binary Logistic), click
Save.
o Check the box for Collinearity diagnostics.
o After running the regression, check VIF values under Coefficients (available if
collinearity is requested). Like linear regression, VIF should be below 5.
4. Adequate Sample Size:
o Make sure you have enough cases, especially for each level of your categorical
variables. SPSS does not automatically test this, so ensure your sample size is
large enough to detect meaningful results (at least 10 cases per predictor).

MULTICOLINEARITY LOGISTIC

Checking Multicollinearity in Linear Regression (for comparison)

If you're specifically interested in checking multicollinearity in a linear regression context, here
are the steps:
1. Run Linear Regression:
o Go to Analyze > Regression > Linear.
o Enter your dependent variable and independent variables.
2. Request Collinearity Diagnostics:
o Click on the Statistics button.
o Check the box for Collinearity Diagnostics.
3. Examine the Output:
o Look for the Collinearity Statistics table in the output.
o VIF Values will be provided here, and you can interpret them in the same way as
described above.
Summary of Steps
 For both logistic and linear regression, the key steps to check for multicollinearity
involve:
1. Running the regression analysis.
2. Requesting collinearity diagnostics to obtain VIF values.
3. Interpreting the VIF values to determine if multicollinearity is a concern.
Important Note
While VIF is a commonly used method to assess multicollinearity, you can also check
correlation matrices or tolerance values as additional methods. Tolerance is calculated as
1/VIF1 / VIF1/VIF, and values below 0.1 may indicate multicollinearity.
In summary, you do not need to switch to linear regression to check for multicollinearity in
logistic regression, as SPSS provides the necessary diagnostics within the logistic regression
analysis itself.
1. Look at the Coefficients (B or Exp(B))
 In logistic regression output, you'll typically see two key values:
o B (log-odds coefficient): This shows the logarithm of the odds of the outcome
happening for each unit increase in the predictor.
o Exp(B) (Odds Ratio): This is the exponentiated version of B and indicates the
odds ratio for the outcome given a one-unit change in the predictor. This value
is easier to interpret.

2. Interpreting the Direction of the Relationship:

 If B (or Exp(B)) is positive:
o The predictor increases the likelihood of the outcome occurring.
o For Exp(B): An odds ratio greater than 1 means that for each unit increase in the
predictor variable, the odds of the outcome occurring increase.
Example: If Exp(B) = 1.5 for age, it means that for every one-year increase in age, the odds of
the outcome (e.g., "success") are 1.5 times greater.
 If B (or Exp(B)) is negative:
o The predictor decreases the likelihood of the outcome occurring.
o For Exp(B): An odds ratio less than 1 means that for each unit increase in the
predictor variable, the odds of the outcome occurring decrease.
Example: If Exp(B) = 0.7 for age, it means that for every one-year increase in age, the odds of
the outcome (e.g., "success") are 0.7 times lower (or 30% less likely).

3. Direction and Significance:

 Significance (p-value < 0.05): If a predictor is significant (p < 0.05), you interpret its
coefficient (B or Exp(B)) as having a real impact on the outcome.
 Non-significant (p-value ≥ 0.05): If the predictor is not significant, you would not make
inferences about its relationship with the outcome, as its effect is not statistically
meaningful.

4. Example of Interpretation:
Imagine you’re predicting whether someone will pass or fail a test (outcome: pass = 1, fail = 0).
The predictors are study time and age.
 If study time has Exp(B) = 2.0, you’d say that for each additional hour of study, the
odds of passing the test are doubled.
 If age has Exp(B) = 0.9, you’d say that for each additional year of age, the odds of
passing the test decrease by 10%.
In summary:
 Positive coefficients (B) or Exp(B) greater than 1: Increase in likelihood of the outcome.
 Negative coefficients (B) or Exp(B) less than 1: Decrease in likelihood of the outcome.

Assumption Checks:
1. Normality:
- Purpose: To verify if your data is normally distributed, a key assumption for parametric tests
like t-tests and ANOVA.

Tests for Normality in SPSS:

- Shapiro-Wilk Test: Recommended for small samples (N < 50).
- Kolmogorov-Smirnov Test: Better for larger samples.

Steps in SPSS:
- Analyze > Descriptive Statistics > Explore.
- Move the variable(s) to the Dependent List.
- Click Plots and check Normality plots with tests.

Interpretation:
- If p > .05: The data is normally distributed.
- If p < .05: Data deviates significantly from normal distribution.

Visual Check:
- Use Q-Q Plot or Histogram (also available under Plots) to visually assess normality.
- Q-Q Plot: Points should fall along the diagonal line if normal

2. Multicollinearity:
- Purpose: To ensure your independent variables are not too highly correlated (important for
regression models).

How to Check in SPSS:

- Analyze > Regression > Linear.
- Move your dependent and independent variables.
- Click Statistics and check Collinearity Diagnostics.

Interpretation:
- Variance Inflation Factor (VIF):
- VIF < 5: Acceptable (closer to 1 is ideal).
- VIF > 10: Indicates problematic multicollinearity.

- Tolerance:
- Tolerance > .2: Acceptable.
- Tolerance < .2: Suggests multicollinearity.

3. Linearity:
- Purpose: Assumes a linear relationship between dependent and independent variables in
regression analysis.
 SPSS Linearity:
- Analyze > Regression > Linear.
- Click on Plots.
- Place ZRESID (Standardized Residuals) on the Y-axis and ZPRED (Standardized Predicted
Values) on the X-axis.
- Click Continue and OK.

Interpretation:
- Look at the scatterplot of residuals vs. predicted values. If the points are randomly dispersed,
the assumption of linearity is met.
- If there’s a clear pattern (e.g., curved), the linearity assumption may be violated.

4. Homoscedasticity:
- Purpose: Assumes that the residuals have constant variance at each level of the independent
variable(s) in regression.

How to Check in SPSS:

- Same steps as checking Linearity (see above), using the scatterplot of residuals vs. predicted
values.

Interpretation:
- If the spread of residuals is roughly the same across all values of predicted scores (random
cloud), homoscedasticity is met.
- If you see a funnel shape or other pattern, heteroscedasticity might be an issue (variance is
not constant).

---

5. Outliers:
- Purpose: Detect if extreme values are unduly influencing your results.

Cook's Distance (for Regression):

- Analyze > Regression > Linear.
- Click Save, then check Cook’s Distance under Distances.

Interpretation:
- Look at Cook’s Distance in your output.
- Cook’s Distance > 1: Indicates an influential outlier.

Boxplots (for detecting outliers in general):

- Analyze > Descriptive Statistics > Explore.
- Under Plots, check Boxplots.
- Outliers will appear as separate dots in the boxplot.
6. Independence of Errors:
- Purpose: In regression, errors (residuals) should be independent from one another.

Durbin-Watson Test (for Autocorrelation):

- Analyze > Regression > Linear.
- Under Statistics, check Durbin-Watson.

Interpretation:
- Durbin-Watson (independence of observations) value should be between 1 and 3.
- Values close to 2 indicate no autocorrelation.
- Values close to 0 or 4 suggest positive or negative autocorrelation, respectively.

7. Effect Size:
- For t-tests:
- Cohen’s d: Small (0.2), Medium (0.5), Large (0.8).
- For correlation: r-value is the effect size.
- For regression: Use R² as effect size (proportion of variance explained).

8. Reporting Key Values:

- T-tests: t(df) = X, p = Y, d = Z.
- Correlation: r(df) = X, p = Y.
- Regression:
- R² = X, F(df1, df2) = X, p = Y, Durbin-Watson = Z.
- Coefficients: B = X, β = Y, p = Z.

9. SPSS Tips:
- Always check Descriptives before running tests.
- Look at Normality (histograms, Q-Q plots, Shapiro-Wilk) to check if assumptions hold.
- Use Transform > Compute to create new variables (e.g., log transformations).
- Use Plots for assumptions testing (Residuals vs Predicted, P-P plots).
 ASSUMPTION VIOLATION (LINEAR)
To check whether the assumptions of linear regression are violated in SPSS, follow these steps
for each assumption:

1. Linearity
- What to Check: The relationship between the independent variables and the dependent
variable should be linear.
- How to Check:
- Run a linear regression analysis:
- Go to Analyze > Regression > Linear.
- Include your dependent variable and independent variables.
- In the Plots menu, select the Standardized Residuals for the Y-axis and Standardized
Predicted Values for the X-axis.
- Look for a random scatter of points around zero; if the plot shows a clear curve, this indicates
a violation of linearity.

2. Independence of Residuals
- What to Check: Residuals (errors) should be independent (no autocorrelation).
- How to Check:
- Conduct the Durbin-Watson test:
- After running your linear regression, check the Model Summary output for the Durbin-
Watson statistic.
- A value close to 2 suggests independence, while values below 1 or above 3 indicate
possible autocorrelation.

3. Homoscedasticity
- What to Check: The residuals should have constant variance at all levels of the independent
variables.
- How to Check:
- In the Plots menu of the linear regression dialog, select ZRESID (standardized residuals) for
the Y-axis and ZPRED (standardized predicted values) for the X-axis.
- Look for a random scatter of points without a distinct pattern. A funnel shape (wider spread of
residuals at higher levels of predicted values) indicates heteroscedasticity.

4. Normality of Residuals
- What to Check: The residuals should be approximately normally distributed.
- How to Check:
- Generate a Histogram of the residuals:
- Save the residuals in the regression output by going to Save in the linear regression dialog
and checking the box for Unstandardized Residuals.
- Then go to Graphs > Legacy Dialogs > Histogram and plot the histogram of the residuals.
- Additionally, you can run the Shapiro-Wilk test:
- Go to Analyze > Descriptive Statistics > Explore.
- Move your residuals variable into the Dependent List and check the option for Normality
plots with tests.
 - If the Shapiro-Wilk test shows a significant p-value (typically p < 0.05), this indicates a
violation of normality.

5. No Multicollinearity
- What to Check: The independent variables should not be too highly correlated with each other.
- How to Check:
- In the linear regression dialog, click on Statistics, and check the box for Collinearity
Diagnostics.
- After running the regression, check the Collinearity Statistics in the output.
- Look for the Variance Inflation Factor (VIF) values:
- A VIF greater than 10 indicates high multicollinearity. As a rule of thumb, VIF values
between 1 and 5 are acceptable.
 Steps to Report Unstandardized Coefficients (B)

1. Identify the Dependent Variable:

o Clearly state the dependent variable that you are analyzing. For example, "The
dependent variable in this analysis is the perceived profile picture rating
(FB_Profile_TOT)."
2. Present the Model Results:
o Include a table or mention the unstandardized coefficients from your SPSS
output. The coefficients will typically be listed in the "Coefficients" table of your
regression output.
o For example, "The unstandardized coefficients (B) for the model are as follows:"

Predictor B (Unstandardized Coefficient)

Constant 1.25
Gender (Male) 0.45
Age 0.10
Extraversion 0.30
Narcissism 0.20

3. Interpret the Coefficients:

o Describe what each coefficient means in relation to the dependent variable:
 Constant (Intercept): This value represents the predicted score of the
dependent variable (FB_Profile_TOT) when all predictors are set to zero.
For example, "The constant value of 1.25 suggests that when all predictor
variables are at their reference levels (e.g., Gender = Female, Age = 0,
Extraversion = 0, Narcissism = 0), the average FB_Profile_TOT rating is
1.25."
 Predictor Variables: Each B value indicates how much the dependent
variable is expected to change when the corresponding independent
variable increases by one unit, holding all other variables constant.
 For instance, "For every one-unit increase in Extraversion, the
FB_Profile_TOT rating is expected to increase by 0.30, controlling
for other factors. This suggests that individuals with higher levels
of extraversion tend to have more favorable profile picture
ratings."
 Similarly, "A one-unit increase in Narcissism is associated with a
0.20 increase in the FB_Profile_TOT rating, indicating a positive
relationship."
4. Provide Context for Significance:
o If relevant, report the statistical significance of these coefficients (usually through
p-values) to indicate whether the effects are statistically significant.
o For example, "The coefficient for Extraversion was statistically significant (B =
0.30, p < .01), suggesting a meaningful impact on the profile picture ratings."
Example Report
Here's a comprehensive example of how you might present this information in a report:

Regression Analysis Results

A multiple linear regression analysis was conducted to predict the perceived profile picture
rating (FB_Profile_TOT) based on gender, age, extraversion, and narcissism. The results of the
regression analysis are summarized in the table below:

Predictor B (Unstandardized Coefficient) p-value

Constant 1.25
Gender (Male) 0.45 0.04
Age 0.10 0.01
Extraversion 0.30 <0.01
Narcissism 0.20 0.02

The unstandardized coefficient for the constant (B = 1.25) indicates that when all predictors are
at their reference levels, the average FB_Profile_TOT rating is expected to be 1.25. For every
one-unit increase in age, the FB_Profile_TOT rating increases by 0.10 (p < 0.01), suggesting
that older individuals tend to receive higher ratings. Additionally, for each unit increase in
extraversion, the rating increases by 0.30 (p < 0.01), highlighting the positive impact of
extraversion on perceived profile picture ratings. Narcissism also positively influences the rating,
with a coefficient of 0.20 (p = 0.02), although its effect is less pronounced than that of
extraversion. The coefficient for gender (B = 0.45, p = 0.04) suggests that males tend to receive
higher ratings compared to females.