README.md

In Ayushi-712/ssclip: SAMPLE SIZE REQUIRED TO TARGET PRECISE ESTIMATES OF PREDICTIVE PERFORMANCE

Sample Size for CLInical Prediction models (ssclip)

The goal of "ssclip" is to calculate minimum sample size for external validation of clinical prediction models to target precise estimates of predictive performance.

There are 6 functions to target different criteria of predictive performance:

1. R2 :The proportion of variance explained.
2. Calibration-in-the-large(CITL) : Agreement between predicted and observed outcome values on average.
3. Calibration slope : agreement between predicted and observed values across the range of predicted values.
4. Residual variances(small multiplicative margin of error (MMOE) around the true value) :To target variance of observed outcome values.
5. Sensitivity : To target model's ability to predict true positives in the dichotomous outcome
6. Specificity : To target model's ability to predict true negatives in the dichotomous outcome

Installation

You can install the development version from GitHub with:

install.packages("devtools")
devtools::install_github("Ayushi-712/ssclip")

Step-by-step guide

Examples (For continuous outcome)

1.Precise estimate of R2val

For Eg.Sample size to target a 95% confidence interval for R2val that has a narrow width of about 0.1 i.e SE of R2val = 0.0255 and expected R2val = 0.5.One can give parameters as R2val= 0.5,width=0.1,alpha=0.05 in the ss_R2val function.

```{r example } library(ssclip) ss_R2val( R2val= 0.5,width=0.1,alpha=0.05 )

#### Output
![Alt desc](https://github.com/Ayushi-712/ssclip/blob/master/Data/ss_R2val.png)
#### for different combinations of parameters one can pass a vector instead of single value.

ss_R2val( R2val= c(0.6,0.9),width=c( 0.1),alpha=c( 0.01,0.05) )

#### Output
![Alt desc](https://github.com/Ayushi-712/ssclip/blob/master/Data/ss_R2val_diff_comb.png)

### 2.Precise estimate of CITL(calibration-in-the-large)
#### For Eg. Sample size to target SE of CITL model of 2.55 (width = 10) at 95% confidence interval , R2 CITL = R2val = 0.5 and variance of the observed Yi = 400.
```{r example }
ss_citl( R2= 0.5,width = 10,alpha = 0.05, varY = 400)

3. Precise estimate of calibration slope

Eg. Sample size to target a 95% confidence interval for 𝜆cal that has a narrow width ≤ 0.2 (eg, if the calibration slope was 1, the confidence interval would be 0.9 to 1.1, confidence intervals derived by 𝜆̂cal ± 1.96SE𝜆̂cal) and expected R2val = 0.5

ss_cal_slope( R2= 0.5,width = 0.2,alpha = 0.05, lambda = 1)

4. Precise estimates of residual variances (small multiplicative margin of error (MMOE) around the true value)

Eg.Sample size for margin of error of within 10% (1.0 <=MMOE <=1.1) of the true value at 95% confidence level.

ss_res_var( max_MOE=1.1, alpha=0.05)

Examples (For dichotomous outcome)

5. To target sensitivity

Eg .Sample size to target 80% sensitivity at 95% confidence level, maximum margin of error 5% for a precision of 95% and 30% of the population with particular disease.

ss_sens( alpha=0.05, se= 0.8, d=0.05,prev=0.3)

6.To target specificity

Eg .Sample size to target 50% specificity at 95% confidence level, maximum margin of error 5% for a precision of 95% and assuming 30% of the population with particular disease.

ss_spec( alpha=0.05, sp= 0.5, d=0.05,prev=0.3)

References

1. Archer L, Snell KIE, Ensor J, Hudda MT, Collins GS, Riley RD. Minimum sample size for external validation of a clinical prediction model with a continuous outcome. Stat Med. 2020;40:133-46.

Ayushi-712/ssclip documentation built on Feb. 4, 2022, 11:50 p.m.