# Test for a qualitative and quantitative difference between groups using IMD and ANOVA, respectively This is the IMD-ANOVA test defined in Webb-Robertson et al. (2010). ```r imd_anova( omicsData, comparisons = NULL, test_method, pval_adjust_a_multcomp = "none", pval_adjust_g_multcomp = "none", pval_adjust_a_fdr = "none", pval_adjust_g_fdr = "none", pval_thresh = 0.05, equal_var = TRUE, model_selection = "auto", parallel = TRUE ) ``` ## Arguments - `omicsData`: pmartR data object of any class, which has a `group_df` attribute created by the `group_designation()` function - `comparisons`: data frame with columns for "Control" and "Test" containing the different comparisons of interest. Comparisons will be made between the Test and the corresponding Control. If left NULL, then all pairwise comparisons are executed. - `test_method`: character string specifying the filter method to use: "combined", "gtest", or "anova". Specifying "combined" implements both the gtest and anova filters. - `pval_adjust_a_multcomp`: character string specifying the type of multiple comparison adjustment to implement for ANOVA tests. Valid options include: "bonferroni", "holm", "tukey", and "dunnett". The default is "none" which corresponds to no p-value adjustment. - `pval_adjust_g_multcomp`: character string specifying the type of multiple comparison adjustment to implement for G-test tests. Valid options include: "bonferroni" and "holm". The default is "none" which corresponds to no p-value adjustment. - `pval_adjust_a_fdr`: character string specifying the type of FDR adjustment to implement for ANOVA tests. Valid options include: "bonferroni", "BH", "BY", and "fdr". The default is "none" which corresponds to no p-value adjustment. - `pval_adjust_g_fdr`: character string specifying the type of FDR adjustment to implement for G-test tests. Valid options include: "bonferroni", "BH", "BY", and "fdr". The default is "none" which corresponds to no p-value adjustment. - `pval_thresh`: numeric p-value threshold, below or equal to which biomolecules are considered differentially expressed. Defaults to 0.05 - `equal_var`: logical; should the variance across groups be assumed equal? - `model_selection`: Character, one of 'full', 'reduced', or 'auto' indicating the model to be used in the ANOVA analysis. The default 'auto' performs an F-test to determine if the full model is necessary. If the F-test is significant, the full model is used, otherwise the reduced model is used. 'full' uses all main effects, covariates, and interactions between main effects. 'reduced' does not consider interactions between main effects (only covariates and marginal main effects). - `parallel`: logical value indicating whether or not to use a "doParallel" loop when running the G-Test with covariates. Defaults to TRUE. ## Returns An object of class 'statRes', which is a data frame containing columns (when relevant based on the test(s) performed) for: e_data cname, group counts, group means, ANOVA p-values, IMD p-values, fold change estimates on the same scale as the data (e.g. log2, log10, etc.), and fold change significance flags (0 = not significant; +1 = significant and positive fold change (ANOVA) or more observations in test group relative to reference group (IMD); -1 = significant and negative fold change (ANOVA) or fewer observations in test group relative to reference group (IMD)) ## Examples ```r library(pmartRdata) # Transform the data mymetab <- edata_transform(omicsData = metab_object, data_scale = "log2") # Group the data by condition mymetab <- group_designation(omicsData = mymetab, main_effects = c("Phenotype")) # Apply the IMD ANOVA filter imdanova_Filt <- imdanova_filter(omicsData = mymetab) mymetab <- applyFilt(filter_object = imdanova_Filt, omicsData = mymetab, min_nonmiss_anova = 2) # Implement IMD ANOVA and compute all pairwise comparisons # (i.e. leave the comparisons argument NULL), with FDR adjustment anova_res <- imd_anova(omicsData = mymetab, test_method = "anova", pval_adjust_a_multcomp = "Holm", pval_adjust_a_fdr = "BY") imd_res <- imd_anova(omicsData = mymetab, test_method = "gtest", pval_adjust_g_multcomp = "bon", pval_adjust_g_fdr = "BY") imd_anova_res <- imd_anova(omicsData = mymetab, test_method = "combined", pval_adjust_a_fdr = "BY", pval_adjust_g_fdr = "BY") imd_anova_res <- imd_anova(omicsData = mymetab, test_method = "combined", pval_adjust_a_multcomp = "bon", pval_adjust_g_multcomp = "bon", pval_adjust_a_fdr = "BY", pval_adjust_g_fdr = "BY") # Two main effects and a covariate mymetab <- group_designation(omicsData = mymetab, main_effects = c("Phenotype", "SecondPhenotype"), covariates = "Characteristic") imd_anova_res <- imd_anova(omicsData = mymetab, test_method = 'comb') # Same but with custom comparisons comp_df <- data.frame( check.names = FALSE, Control = c("Phenotype1", "A"), Test = c("Phenotype2", "B") ) custom_comps_res <- imd_anova(omicsData = mymetab, comparisons = comp_df, test_method = "combined") ``` ## References Webb-Robertson, Bobbie-Jo M., et al. "Combined statistical analyses of peptide intensities and peptide occurrences improves identification of significant peptides from MS-based proteomics data." Journal of proteome research 9.11 (2010): 5748-5756. ## Author(s) Bryan Stanfill, Kelly Stratton

imd_anova function