New/Changed job control qualifiers

Changes to PREDICT

The following are new qualifiers for the PREDICT directive
!ESTIMATE [coef]     This PREDICT qualifiecr instructs ASReml to form the full variance/covariance matrix for the predicted values and then calculate a series of linear functions across the set of predictions. For example, if we have a factor SNP with 3 classes AA, AB, BB in that order,
PREDICT SNP !ESTIMATE -1 0 1 -1 2 -1
would report 2 contrasts over the 3 predicted values.

!PAC [t]     (Predict Actual Coefficients) is used instead of !ONLYUSE t when t is a design function of a variable (like leg(), mbf() or spl()), and the prediction pertains to the actual coefficients, not the function of the coefficients. For example,
!MBF mbf(times,12) MyZeds.txt !SKIP 1 !KEY 1 !RENAME Zeds
accel ~ mu times !r Zeds
PREDICT Zeds 1 2 3 4 5 6 7 8 9 10 11 12 !PAC Zeds !VPV
where times is a covariate and MyZeds.txt has a leading key variable listing the unique times followed by 12 base variables. With !PAC also specified, the (default) predict points 1:12 (in the PREDICT line) index the base variables rather than being covariate values for times. Job output qualifier !CINV Zeds is another way of getting the variance matrix for the 12 effects.

!SPLIT This PREDICT qualifier causes the predict output from each prediction to be split into multiple files for easier parsing. By default, he default .pvs file holds all explanatory text along with the prediction tables. With !SPLIT, the predicted value lines are reported in an basenamePTi.txt (.csv) file in a form easily loaded into Excel or R. All explanatory text associated with the predict table are written to basenamePHi.txt file where ii is the index of the PREDICT statement.

!TDIFF This PREDICT qualifier requests t-statistics be printed for all combinations of predicted values. It creates a pairwise table of t-statistics among all predicted values (assuming the number of cells predicted is not ridiculously large). ASReml now reports Bonferroni and Tukey critical values for assessing the significance of the t-statistics in some common situations. For a 1-way table, the denominator degrees of freedom for the factor need to have been calculated for the Wald F-table. For a 2-way table, the variance model needs to be simple, such as a split-plot, in which stratum var

Changes to TABULATE

!PRINTALL     This qualifier reports empty cells in the tabulation. By default, such cells are not reported. This is useful for identifying where empty cells appear in the class list. 10.2
  • Back to What's New

    Return to index