Input to BootCI is a tab-delimited text file of the sort easily exported from statistical packages like SPSS or spreadsheet programs like Excel. The first line consists of variable labels separated by tabs. Successive lines contain values for these variables, again separated by tabs. If N is the number of cases, the file contains N+1 lines.
The number of variables is limited to 10 including the outcome variable, primarily because Algina and colleagues in their simulation study investigated the coverage (best is 100%; see Algina et al., 2007, for definitions) for k = 3, 6, and 9 predictor variables for sample sizes of 50, 100, 150, and 200 (see Table 1). Clearly, more predictor variables require larger sample sizes, even for percentile bootstrap CIs.
Table 1. Bootstrap CI Coverage with 3, 6, and 9 predictor variables and various sample sizes.
|# of predictors||N = 50||N = 100||N = 150||N = 200|
|k = 3||100%||100%||100%||100%|
|k = 6||73%||100%||100%||100%|
|k = 9||50%||70%||90%||95%|
Note. From Algina et al., 2007; see article for details.
After opening a data file with eight variable, selecting options, and picking “CI for sr2” (the wizard icon), the BootCI output looks like this.
The DV is XLang. The base model includes pWMP, to which FC (the IV) is added. Alone pWPM accounts for 11% of XLang’s variance. FC accounts for an additional 16%. Its bootstrap CIs (CI sampled) are [.05, .31].