Skip to main content

Table 5 Standard error before and after batch adjustment

From: Propensity scores as a novel method to guide sample allocation and minimize batch effects during the design of high throughput experiments

 

No adjustment

ComBat adjustment

Regression adjustment

 

Mean

Min

Mean

Min

Mean

Min

Case (\({\upbeta }_{1})\) | True Value = 1.40E-01

Randomization Condition

2.37E-01

1.48E-01

1.42E-01

1.37E-01

1.50E-01

1.45E-01

Stratified Randomization Condition

2.38E-01

1.47E-01

1.42E-01

1.39E-01

1.49E-01

1.45E-01

Optimal Condition

2.41E-01

1.47E-01

1.42E-01

1.41E-01

1.48E-01

1.47E-01

Age (\({\beta }_{2})\) | True Value = 5.09E-03

Randomization Condition

8.59E-03

5.35E-03

5.13E-03

4.97E-03

5.44E-03

5.21E-03

Stratified Randomization Condition

8.64E-03

5.34E-03

5.13E-03

5.02E-03

5.44E-03

5.26E-03

Optimal Condition

8.73E-03

5.35E-03

5.14E-03

5.13E-03

5.35E-03

5.34E-03

HbA1c (\({\beta }_{3})\) | True Value = 1.34E-01

Randomization Condition

2.26E-01

1.40E-01

1.35E-01

1.31E-01

1.43E-01

1.37E-01

Stratified Randomization Condition

2.27E-01

1.40E-01

1.35E-01

1.32E-01

1.43E-01

1.38E-01

Optimal Condition

2.29E-01

1.40E-01

1.35E-01

1.35E-01

1.41E-01

1.40E-01

  1. Within each experimental iteration, standard error calculated as the average standard error across the 10,000 most variable genes in the dataset
  2. Mean mean of average standard error across all experimental iterations; Min minimum of average standard error across all experimental iterations
  3. Red highlighting identifies standard error values less than standard error values estimated in the ‘true’ gene expression dataset (before batch effects were added). The emboldened values represent standard error estimates that are less than the corresponding true value