Robert’s Working Blog

1. it needs to divide from 99%–10% to build CI

2. it needs to build crossing section matrix,which stores crossing section for each patient at 40 times.

3. it needs to consider the average corssing section value and total number of corssing sections

4. it’s best to have choice to display 99%–10% seperately

5. we need to divide patients into <36, 36-37, 37-38, 38-39… to see crossing section and aim to combine risk factor to find some patterns (trajectory behavoir: parallel, crossing…).

6. It nees to build 99%–10% CI contour for  “preterm birth”

He suggests to have a talk with “Leigh Johnson”

http://www.themis-researcher-profile.unimelb.edu.au/researcher/person15352.html

For PC, talking to D. Botterham (IT Manager). 

Today:

1. programming on 4 sectors 95% CI 

Today works on Peter’s task:

1. Reading papers 

Meeting with David:

1. Keep working on “Using 95% CI to predict” but need to cut into 8 detailed sections to calculate the CI.

2. Building 99%, 95%, 90%, 80%, 70%, 60%, 50%… 

3. Once CI is finished, building position matrix for each patients (Horizon: 1–>40 weeks. Vertical: PSN1–>PSN420).

4. Predict based on the section of each patient’s location. 

David also mentions individual trajectory analysis, for each patient, loading on data point, draw a circle for error torrent then see how many patients crossing in this area;

loading the seond data points, draw a circel for error torrent then see how many patients crossing in this area; The general idea is to reduce the percentage of patients’ to reduce the final area of patient’s possible going then using this area to predict the percentage.

Meet Bruce about Confidence Interval and he suggests to:

1. calculate the distance to the mean value point

2. Then apply 95% confidence interval (CI), which is the same as David’s suggestion. It’s perfect.

There are several tasks after the meeting with Roger and David:

1. Building different matrix for each individual trajectory and average trajectory (individual - average) 

2. Alyways use “distance” between individual and average to define “parallel, crossing”

3. Calculate the “standard deviation” for each part (left part of crossing and right part)

4. Try to define “crossing” –> distance=0 and STD is a large number

“parallel” –> distance=fixed number and STD is 0 

1. working on filter data between “riskfactor” and “rawdata”

2. working on adding “twins” data in “rawdata”, but there is no “twins” data for ultrasound

3. working on adding “Apply” button on “riskfactor” panel

4. reading Roger’s paper 

(1). Reading webpages from: http://en.wikipedia.org/wiki/Premature_birth

shows some risk factors for PTB:

1. Previous history of preterm birth

2. Multiple pregnancies

3. Certain chronic disease such as high blood pressure, kidney disease and diabetes.

4. Substance abuse of tobacco, alcohol and other drugs

5. Women under 18 or over 35 are at a higher risk for premature birth

We should make full use of our data to validate to see whether we can find some useful patterns from these risk factors.

(2). Reading webpages from: http://en.wikipedia.org/wiki/Birth_weight

Categories of growth retardation: Symmetrical growth retardation and Asymmetrical growth retardation

We should also try to see whether our data has these patterns or not. “BPD” is the parameter, we should focus on.