Bayesian Statistics:
←Older revision
Revision as of 01:48, 23 April 2015
Line 52:
Line 52:
<!-- Instructions: Write a medium-length summary (~10 - 20 sentences) of how BME100 tested patients for the disease-associated SNP. Describe (A) the division of labor (e.g., 34 teams of 6 students each diagnosed 68 patients total...), (B) things that were done to prevent error, such as the number of replicates per patient, PCR controls, ImageJ calibration controls, and the number of drop images that were used for the ImageJ calculations (per unique PCR sample), and (C) the class's final data from the BME100_fa2014_PCRResults spreadsheet (successful conclusions, inconclusive results, blank data). -->
<!-- Instructions: Write a medium-length summary (~10 - 20 sentences) of how BME100 tested patients for the disease-associated SNP. Describe (A) the division of labor (e.g., 34 teams of 6 students each diagnosed 68 patients total...), (B) things that were done to prevent error, such as the number of replicates per patient, PCR controls, ImageJ calibration controls, and the number of drop images that were used for the ImageJ calculations (per unique PCR sample), and (C) the class's final data from the BME100_fa2014_PCRResults spreadsheet (successful conclusions, inconclusive results, blank data). -->
-
+
The labor of testing each of the 16 patients was divided among 8 lab groups, each made up of 4-5 BME 100 students. Each group was given two unidentified samples to test for Type I Diabetes using SYBR Green 1 as an indicator of positive or negative. Positive samples were shown to glow green when photographed while negative samples remained relatively inert and somewhat colorless. Three replicate tests were performed per patient, and the results from these tests factored in the ultimate diagnosis for that patient. Any questionable result was labelled "inconclusive," and discounted from the whole of the data. Data that contained abundant error that through off experimental results was disregarded, as per instruction from the instructor that lab period.
'''What Bayes Statistics Imply about This Diagnostic Approach'''
'''What Bayes Statistics Imply about This Diagnostic Approach'''
+
+
Bayesian Statistics Equation (Used in all Calculations)
+
+
P(A|B) = {[P(B|A) * P(A)] / P(B)} * 100%
+
+
'''Calculation 1'''
+
+
Calculation 1 involved determining the probability that any given sample has the Type I Diabetes DNA sequence given a positive diagnostic signal.
+
+
In which:
+
+
*A: frequency/fraction of diabetes-positive conclusions
+
*B: frequency/fraction of positive PCR reactions
+
*P(B|A) = The frequency/fraction of positive PCR reactions given diabetes positive conclusion.
+
+
+
'''Calculation 2'''
+
+
Calculation 2 involved determining the probability that any given sample does not have the Type I Diabetes DNA sequence given a negative diagnostic signal.
+
+
In which:
+
+
*A: frequency/fraction of diabetes-negative conclusions
+
*B: frequency/fraction of negative PCR reactions
+
*P(B|A) = The frequency/fraction of negative PCR reactions given diabetes negative conclusion.
+
+
+
'''Calculation 3'''
+
+
Calculation 3 involved determining the probability of a patient developing Type I Diabetes given a Diabetes DNA sequence.
+
+
In which:
+
+
*A: frequency/fraction of diabetes "yes" diagnosis
+
*B: frequency/fraction of positive "pos" DNA test conclusion
+
*P(B|A) = The frequency/fraction of positive (pos) given diabetes diagnosis "yes"
+
+
+
'''Calculation 4'''
+
+
Calculation 4 involved determining the probability that a patient will not develop Type I Diabetes given a non-Diabetes DNA sequence.
+
+
*A: frequency/fraction of diabetes "no" diagnosis
+
*B: frequency/fraction of negative "neg" DNA test conclusion
+
*P(B|A): The frequency/fraction of negative (neg) given diabetes diagnosis "no"
<!-- Instruction 1: In your own words, discuss what the results for calculations 1 and 2 imply about the reliability of the individual PCR replicates for concluding that a person has the disease SNP or not. Please do NOT type the actual numerical values here. Just refer to the Bayes values as being "close to 1.00 (100%)" or "very small." Discuss at least three possible sources of human or machine/device error that could have occurred during the PCR & detection steps that could have affected the Bayes values in a negative way. -->
<!-- Instruction 1: In your own words, discuss what the results for calculations 1 and 2 imply about the reliability of the individual PCR replicates for concluding that a person has the disease SNP or not. Please do NOT type the actual numerical values here. Just refer to the Bayes values as being "close to 1.00 (100%)" or "very small." Discuss at least three possible sources of human or machine/device error that could have occurred during the PCR & detection steps that could have affected the Bayes values in a negative way. -->
+
+
<ol>
+
+
<li> yada </li>
+
<!-- Instruction 1: In your own words, discuss what the results for calculations 3 and 4 imply about the reliability of PCR for *predicting the development disease* (referred to as "diagnosis"). Please do NOT type the actual numerical values here. Just refer to the Bayes values as being "close to 1.00 (100%)" or "very small." -->
<!-- Instruction 1: In your own words, discuss what the results for calculations 3 and 4 imply about the reliability of PCR for *predicting the development disease* (referred to as "diagnosis"). Please do NOT type the actual numerical values here. Just refer to the Bayes values as being "close to 1.00 (100%)" or "very small." -->
+
+
<li> dada </li>
+
+
</ol>
==Computer-Aided Design==
==Computer-Aided Design==