CALL td_analyze (
'ranktest',
'required_parameter_list [ optional_parameter; [...] ]'
);
- required_parameter_list
database = input_database_name;
tablename = input_table_name;
- optional_parameter
{ blockcolumn = column_name |
columnofinterest = column_name |
columns = column_name [,...] |
fallback = { true | false } |
firstcolumn = column_name |
gensqlonly = { true | false } |
groupby = column_name [,...] |
includezero = { true | false } |
independent = { true | false } |
multiset = { true | false } |
outputdatabase = output_database_name |
outputtablename = output_table_name |
overwrite = { true | false } |
secondcolumn = column_name |
singletail = { true | false } |
statsdatabase = stats_database_name |
teststyle = { ms | wilcoxon | friedman } |
thresholdprobability = threshold_prob |
treatmentcolumn = column_name
}
Syntax Elements
- database
- The database containing the table to analyze.
- tablename
- The table containing the columns to analyze. It must reside in the database indicated by the database parameter.
- blockcolumn
- [Optional] Only if teststyle=friedman, the input table column that represents blocks.
When pairing blockcolumn and treatmentcolumn values, unequal cell counts can cause a division-by-zero error.
- columnofinterest
- [Optional] Only if teststyle=ms, the input table column that represents the dependent variable. If nonnumeric, it is ranked alphanumerically.
- columns
- [Optional] Only if teststyle=ms, the input table columns that represent the independent variables.
- fallback
- [Optional] Whether to give the output table the FALLBACK attribute.
- Default: false
- firstcolumn
- [Optional] Only if teststyle=wilcoxon, the input table column that represents the first sample variable.
- gensqlonly
- [Optional] True returns the SQL for the function as a result set but does not run it.
- False runs the SQL for the function but does not return it as a result set.
- Disallowed with teststyle=ms.
- Default: false
- groupby
- [Optional] The input table columns for which to separately analyze each value or combination of values.
- Default behavior: Input is not grouped.
- includezero
- [Optional] Only if teststyle=wilcoxon, whether to include cases with zero differences in the positive count.
- Default: false (discards cases with zero differences)
- independent
- [Optional] Only if teststyle=ms, whether to consider each requested variable individually (a variation of the Mann-Whitney or Kruskal-Wallis test), performing a series of independent tests, rather than in combination.
- Default: false
- multiset
- [Optional] Whether to give the output table the MULTISET attribute.
- Default: false
- outputdatabase
- [Optional] The name of the database to contain the analysis results table.
- outputtablename
- [Optional] The name of the table to store the analysis results.
- Default behavior: Function returns results as a result set but does not create an output table.
- secondcolumn
- [Optional] Only if teststyle=wilcoxon, the input table column that represents the second sample variable.
- singletail
- [Optional] Only if teststyle=ms or teststyle=wilcoxon, whether the test is single-tailed.
- If the Mann-Whitney test becomes a Kruskall-Wallis test, this option is invalid.
- Default: false (test is two-tailed)
- statsdatabase
- [Optional] The database where the statistical test metadata tables are installed.
- Default behavior: The function searches the source database for these tables.
- teststyle
- [Optional] One of the following test styles:
Option |
Description |
mw (default) |
Mann-Whitney test If no independent variable has more than two distinct values, Kruskal-Wallis test otherwise. |
wilcoxon |
Wilcoxen test. |
friedman |
Friedman test with Kendall's Coefficient of Concordance and Spearmans' Rho. |
- thresholdprobability
- [Optional] The threshold (alpha) probability, below which the null hypothesis is rejected.
- Default: 0.05
- treatmentcolumn
- [Optional] Only if teststyle=friedman, the column that represents the independent categorical variable.
When pairing blockcolumn and treatmentcolumn values, unequal cell counts can cause a division-by-zero error.