Difference between revisions of "MSymond1 Week 9"
(→Data & Files: added slide) |
(finished conclusion) |
||
| Line 36: | Line 36: | ||
[[media:Symonds_BIOL367_S24_sample_p-value_slide.pdf]] | [[media:Symonds_BIOL367_S24_sample_p-value_slide.pdf]] | ||
| − | + | ==Conclusion== | |
| − | + | This experiment gives many implications to the viability of using p-values from ANOVA tests. Several of the p values that came out as less than .05 from the ANOVA test did not have a p-value of less than .05 for the Bonferroni test or the B-H test. Not to mention that the p-value being less than .05 is an arbitrary comparison to begin with. With a sample size this large, it is very important to understand that there will be many times the p values seem to indicate a significant result when there in fact is not one. The Bonferroni tests and B-H tests also yielded far less p values that were less than .05. It is very interesting that this has such large implications for the field of biology and most of the sciences, but this has never been taught to the experimenter before in this study. | |
| − | |||
| − | |||
| − | |||
| − | |||
| − | |||
| − | |||
| − | |||
| − | |||
| − | |||
| − | |||
| − | |||
| − | |||
{{Template:MSymond1}} | {{Template:MSymond1}} | ||
Revision as of 20:15, 20 March 2024
Contents
Purpose
This lab was conducted in Microsoft Excel to analyze a microarray dataset. Each group in class was given a different dataset for a different strain of data. The strain used in the present study was the wild type. The ANOVA tests done on the dataset determined which genes had a p value of less than .05, and they were then further analyzed further to discover that an ANOVA test alone is not enough to determine if there is a statistically significant change in the genes over time.
Methods & Results
ANOVA
The data was imported into Microsoft Excel, an ANOVA test was run on every gene (6189) in the data set by first calculating the average of each data point for each time interval (15, 30, 60, 90, 120). Then the sum of squares for each time interval was also calculated using the syntax provided by the lab protocol. The Fstat was then calculated for the full dataset, which then allowed to calculate the p-value for each gene. Once the p-values were all calculated, they were then modified to correct for the multiple testing problem.
Bonferroni and p-value correction
To calculate the Bonferroni p-value, the original p values were all multiplied by 6189. Then, in another column, the Bonferroni p values were all either changed to 1 if they were greater than 1, or they were reported as their Bonferroni p-values if they were less than 1.
Benjamini & Hochberg p-value correction
A new worksheet was created in Excel with all of the raw p-values, the p-values were ranked from least to greatest in this new worksheet. To calculate the new p-values, the p-values were multiplied by 6189 again, and then they were divided by their rank of the p-values. The same process was repeated from the Bonferroni calculations in which they were only reported if they were less than 1.
Sanity Check
- NSR1
- Unadjusted p-value: 2.86939E-10
- Bonferroni p-value: 1.77586E-06
- B-H p-value: 0.111519036
- Average log
- 15: 3.279225
- 30: 3.621
- 60: 3.526525
- 90: -2.04985
- 120: -0.60622
- According to NSR1's unadjusted p-value, and its Bonferroni p-value, it does undergo change in expression during the experiment, but not according to the B-H p-value.
- MSN1
- Unadjusted p-value: 0.563798852
- Bonferroni p-value: 3489.351093
- B-H p-value: 0.679258535
- Average log
- 15: 0.1076
- 30: -0.46192
- 60: -0.47075
- 90: 0.16805
- 120: -0.18418
- According to all 3 p-values, it does not undergo change in expression during the experiment.
Data & Files
media:Symonds_Miller_BIOL367_S24_microarray-data_wt.xlsx.zip media:Symonds_BIOL367_S24_sample_p-value_slide.pdf
Conclusion
This experiment gives many implications to the viability of using p-values from ANOVA tests. Several of the p values that came out as less than .05 from the ANOVA test did not have a p-value of less than .05 for the Bonferroni test or the B-H test. Not to mention that the p-value being less than .05 is an arbitrary comparison to begin with. With a sample size this large, it is very important to understand that there will be many times the p values seem to indicate a significant result when there in fact is not one. The Bonferroni tests and B-H tests also yielded far less p values that were less than .05. It is very interesting that this has such large implications for the field of biology and most of the sciences, but this has never been taught to the experimenter before in this study.
User Page
Assignment Pages
Individual Journal Pages
- MSymond1 Week 1
- MSymond1 Week 2
- MSymond1 KMill104 Week 3
- NeMO_Week4
- MSymond1 Week 5
- MSymond1 Week 6
- MSymond1 Week 8
- MSymond1 Week 9
- MSymond1 Week 10
- MSymond1 Week 12
- MSymond1 Week 13
- MSymond1 Week 15
