Bootstrap validation
Take multiple random samples from a dataset and rerun statistical analyses
The bootstrap is, by far, the most prevalent method for validating statistical findings. Random samples (1000's of them, if you want) of your dataset are taken, statistical analyses are run on each random sample, and a 95% bootstrap confidence interval for the primary finding is generated. If the bootstrap confidence intervals are relatively narrow, then researchers can assume that the findings are valid.
Bootstrap validation in SPSS (simple random sampling method)
The bootstrap validation technique is available when conducting certain statistics in SPSS Version 21.* These statistics include one-sample t-test, independent samples t-test, one-way ANOVA, Fisher's Exact test, chi-square, odds ratio, relative risk, McNemar's test, repeated-measures t-test, phi-coefficient, point biserial, rank biserial, biserial, Spearman's rho, Pearson's r, logistic regression, multinomial logistic regression, proportional odds regression, multiple regression, fixed-effects ANOVA, ANCOVA, MANOVA, MANCOVA, Poisson regression, and negative binomial regression.
Here are the steps for conducting a bootstrap with a simple random sampling technique in SPSS:
1. In the SPSS window interface for the statistic being conducted, click on the Bootstrap... button.
2. Click on the box for Perform bootstrapping to select it. (Leave the Number of samples: box as 1,000 and the Sampling table as Simple.)
3. Click Continue.
4. In the Output file interface, look for the table entitled "Bootstrap for ..." towards the bottom of the output.
5. For mean differences, look in the Mean Difference and 95% Confidence Interval columns of the table. The bootstrap 95% confidence interval of the mean difference is the primary inference yielded from the bootstrap analysis.
6. For differences in proportions, look in the Value and 95% Confidence Interval columns of the table. The bootstrap 95% confidence interval of the value is the primary inference yielded from the bootstrap analysis.
7. For correlations, look in the Correlation and 95% Confidence Interval rows of the table. The bootstrap 95% confidence interval of the Pearson correlation coefficient is the primary inference yielded from the bootstrap analysis.
8. For beta coefficients, look in the B and 95% Confidence Interval columns of the table. The bootstrap 95% confidence interval of the beta coefficient is the primary inference yielded from the bootstrap analysis.
Here are the steps for conducting a bootstrap with a simple random sampling technique in SPSS:
1. In the SPSS window interface for the statistic being conducted, click on the Bootstrap... button.
2. Click on the box for Perform bootstrapping to select it. (Leave the Number of samples: box as 1,000 and the Sampling table as Simple.)
3. Click Continue.
4. In the Output file interface, look for the table entitled "Bootstrap for ..." towards the bottom of the output.
5. For mean differences, look in the Mean Difference and 95% Confidence Interval columns of the table. The bootstrap 95% confidence interval of the mean difference is the primary inference yielded from the bootstrap analysis.
6. For differences in proportions, look in the Value and 95% Confidence Interval columns of the table. The bootstrap 95% confidence interval of the value is the primary inference yielded from the bootstrap analysis.
7. For correlations, look in the Correlation and 95% Confidence Interval rows of the table. The bootstrap 95% confidence interval of the Pearson correlation coefficient is the primary inference yielded from the bootstrap analysis.
8. For beta coefficients, look in the B and 95% Confidence Interval columns of the table. The bootstrap 95% confidence interval of the beta coefficient is the primary inference yielded from the bootstrap analysis.
Bootstrap validation in SPSS (stratified random sampling method)
There is an even more powerful bootstrap methodology available to you in SPSS. The stratified random sampling method allows researchers to randomly sample from different strata of predictor, confounding, or demographic variables, based on what the research questions require in order to conduct sensitivity and subgroup analyses. It is available for all of the aforementioned statistical tests available in SPSS.
Here are the steps for conducting a bootstrap with a stratified random sampling technique in SPSS:
1. In the SPSS window interface for the statistic being conducted, click on the Bootstrap... button.
2. Click on the box for Perform bootstrapping to select it. (Leave the Number of samples: box as 1,000).
3. In the Sampling table, click on the Stratified button to select it.
4. Based on the research question, the specific stratum (predictor, confounding, or demographic variable), or the sampling need, choose which variable(s) in the Variables: box that are to be randomly sampled from in the dataset.
5. Click Continue.
6. Click OK.
7. In the Output file interface, look for the table entitled "Bootstrap for ..." towards the bottom of the output.
8. For mean differences, look in the Mean Difference and 95% Confidence Interval columns of the table. The bootstrap 95% confidence interval of the mean difference is the primary inference yielded from the bootstrap analysis.
9. For differences in proportions, unadjusted odds ratios, adjusted odds ratios, or relative risk, look in the Value and 95% Confidence Interval columns of the table. The bootstrap 95% confidence interval of the value is the primary inference yielded from the bootstrap analysis.
10. For correlations, look in the Correlation and 95% Confidence Interval rows of the table. The bootstrap 95% confidence interval of the Pearson correlation coefficient is the primary inference yielded from the bootstrap analysis.
11. For beta coefficients, look in the B and 95% Confidence Interval columns of the table. The bootstrap 95% confidence interval of the beta coefficient is the primary inference yielded from the bootstrap analysis.
Here are the steps for conducting a bootstrap with a stratified random sampling technique in SPSS:
1. In the SPSS window interface for the statistic being conducted, click on the Bootstrap... button.
2. Click on the box for Perform bootstrapping to select it. (Leave the Number of samples: box as 1,000).
3. In the Sampling table, click on the Stratified button to select it.
4. Based on the research question, the specific stratum (predictor, confounding, or demographic variable), or the sampling need, choose which variable(s) in the Variables: box that are to be randomly sampled from in the dataset.
5. Click Continue.
6. Click OK.
7. In the Output file interface, look for the table entitled "Bootstrap for ..." towards the bottom of the output.
8. For mean differences, look in the Mean Difference and 95% Confidence Interval columns of the table. The bootstrap 95% confidence interval of the mean difference is the primary inference yielded from the bootstrap analysis.
9. For differences in proportions, unadjusted odds ratios, adjusted odds ratios, or relative risk, look in the Value and 95% Confidence Interval columns of the table. The bootstrap 95% confidence interval of the value is the primary inference yielded from the bootstrap analysis.
10. For correlations, look in the Correlation and 95% Confidence Interval rows of the table. The bootstrap 95% confidence interval of the Pearson correlation coefficient is the primary inference yielded from the bootstrap analysis.
11. For beta coefficients, look in the B and 95% Confidence Interval columns of the table. The bootstrap 95% confidence interval of the beta coefficient is the primary inference yielded from the bootstrap analysis.
Statistician For Hire
DO YOU NEED TO HIRE A STATISTICIAN?
Eric Heidel, Ph.D. will provide statistical consulting for your research study at $100/hour. Secure checkout is available with PayPal, Stripe, Venmo, and Zelle.
- Statistical Analysis
- Sample Size Calculations
- Diagnostic Testing and Epidemiological Calculations
- Psychometrics
*SPSS Version 21 (Armonk, NY: IBM Corp.).