From 28b7e54157eeb3ed94111026b1b958a5c17b547c Mon Sep 17 00:00:00 2001 From: msrosenberg Date: Mon, 14 Nov 2022 16:42:24 -0500 Subject: [PATCH 1/2] typo in help file --- resources/metawin_help.html | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/resources/metawin_help.html b/resources/metawin_help.html index aff909e..e2b5385 100644 --- a/resources/metawin_help.html +++ b/resources/metawin_help.html @@ -1054,7 +1054,7 @@

Running a Rank Correlation Analysis

Additionally, there is an option to perform the rank correlation using Kendall's τ (Kendall 1938) or Spearman's ρ (Spearman 1904). Because the distribution of these metrics can be complicated, particularly when the - number of studies is low, MetaWin autmoatically uses a randomization + number of studies is low, MetaWin automatically uses a randomization procedure to thest their significance, with the number of iterations user-specifiable.

From ba149aced04faa6ed9f203438a028f4ac6f8c7f5 Mon Sep 17 00:00:00 2001 From: msrosenberg Date: Mon, 14 Nov 2022 17:13:03 -0500 Subject: [PATCH 2/2] progress bar Added a progress indicator for resampling tests --- src/MetaWinAnalysis.py | 30 +++-- src/MetaWinAnalysisFunctions.py | 187 ++++++++++++++++++++++++++++---- src/MetaWinLanguage.py | 3 + src/MetaWinWidgets.py | 14 ++- 4 files changed, 197 insertions(+), 37 deletions(-) diff --git a/src/MetaWinAnalysis.py b/src/MetaWinAnalysis.py index f6ab1d7..5eada4c 100644 --- a/src/MetaWinAnalysis.py +++ b/src/MetaWinAnalysis.py @@ -1786,7 +1786,7 @@ def add_resampling_options_to_dialog(sender, test_model: bool = False): def do_meta_analysis(data, options, decimal_places: int = 4, alpha: float = 0.05, tree: Optional = None, - norm_ci: bool = True): + norm_ci: bool = True, sender=None): """ primary function controlling the execution of an analysis @@ -1799,42 +1799,48 @@ def do_meta_analysis(data, options, decimal_places: int = 4, alpha: float = 0.05 output_blocks.extend(output) if options.structure == SIMPLE_MA: (output, figure, chart_data, analysis_values, - citations) = MetaWinAnalysisFunctions.simple_meta_analysis(data, options, decimal_places, alpha, norm_ci) + citations) = MetaWinAnalysisFunctions.simple_meta_analysis(data, options, decimal_places, alpha, norm_ci, + sender=sender) elif options.structure == GROUPED_MA: (output, figure, chart_data, analysis_values, - citations) = MetaWinAnalysisFunctions.grouped_meta_analysis(data, options, decimal_places, alpha, norm_ci) + citations) = MetaWinAnalysisFunctions.grouped_meta_analysis(data, options, decimal_places, alpha, norm_ci, + sender=sender) elif options.structure == CUMULATIVE_MA: - output, figure, chart_data = MetaWinAnalysisFunctions.cumulative_meta_analysis(data, options, - decimal_places, alpha, norm_ci) + output, figure, chart_data = MetaWinAnalysisFunctions.cumulative_meta_analysis(data, options, decimal_places, + alpha, norm_ci, sender=sender) analysis_values = None citations = [] elif options.structure == REGRESSION_MA: (output, figure, chart_data, analysis_values, - citations) = MetaWinAnalysisFunctions.regression_meta_analysis(data, options, decimal_places, alpha, norm_ci) + citations) = MetaWinAnalysisFunctions.regression_meta_analysis(data, options, decimal_places, alpha, norm_ci, + sender=sender) elif options.structure == COMPLEX_MA: output, analysis_values, citations = MetaWinAnalysisFunctions.complex_meta_analysis(data, options, decimal_places, alpha, - norm_ci) + norm_ci, sender=sender) figure = None chart_data = None elif options.structure == NESTED_MA: (output, figure, chart_data, analysis_values, - citations) = MetaWinAnalysisFunctions.nested_meta_analysis(data, options, decimal_places, alpha, norm_ci) + citations) = MetaWinAnalysisFunctions.nested_meta_analysis(data, options, decimal_places, alpha, norm_ci, + sender=sender) elif options.structure == TRIM_FILL: (output, figure, chart_data, analysis_values, citations) = MetaWinAnalysisFunctions.trim_and_fill_analysis(data, options, decimal_places, alpha, norm_ci) elif options.structure == JACKKNIFE: (output, figure, chart_data, - citations) = MetaWinAnalysisFunctions.jackknife_meta_analysis(data, options, decimal_places, alpha, norm_ci) + citations) = MetaWinAnalysisFunctions.jackknife_meta_analysis(data, options, decimal_places, alpha, norm_ci, + sender=sender) analysis_values = None elif options.structure == PHYLOGENETIC_MA: output, citations = MetaWinAnalysisFunctions.phylogenetic_meta_analysis(data, options, tree, decimal_places, - alpha, norm_ci) + alpha, norm_ci, sender=sender) analysis_values = None figure = None chart_data = None elif options.structure == RANKCOR: - output, citations = MetaWinAnalysisFunctions.rank_correlation_analysis(data, options, decimal_places) + output, citations = MetaWinAnalysisFunctions.rank_correlation_analysis(data, options, decimal_places, + sender=sender) figure = None chart_data = None analysis_values = None @@ -1904,7 +1910,7 @@ def meta_analysis(sender, data, last_effect, last_var, decimal_places: int = 4, if meta_analysis_options.structure is not None: output, figure, chart_data, _ = do_meta_analysis(data, meta_analysis_options, decimal_places, alpha, tree, - norm_ci) + norm_ci, sender=sender) sender.last_effect = meta_analysis_options.effect_data sender.last_var = meta_analysis_options.effect_vars return output, figure, chart_data diff --git a/src/MetaWinAnalysisFunctions.py b/src/MetaWinAnalysisFunctions.py index d7e0c82..ab2eb4d 100644 --- a/src/MetaWinAnalysisFunctions.py +++ b/src/MetaWinAnalysisFunctions.py @@ -16,6 +16,7 @@ from MetaWinUtils import create_output_table, inline_float, interval_to_str, get_citation, exponential_label, \ prob_z_score import MetaWinCharts +import MetaWinWidgets from MetaWinLanguage import get_text @@ -299,13 +300,20 @@ def median_effect(e: numpy.array, w: numpy.array): return tmp_data[i, 0] -def bootstrap_means(bootstrap_n, boot_data, obs_mean, pooled_var, random_effects: bool = False, alpha: float = 0.05): +def bootstrap_means(bootstrap_n, boot_data, obs_mean, pooled_var, random_effects: bool = False, alpha: float = 0.05, + progress_bar=None): """ conduct a bootstrap test to create confidence intervals around mean effect size """ if bootstrap_n is not None: rng = numpy.random.default_rng() all_means = [obs_mean] + + # if sender is not None: + # progress = MetaWinWidgets.progress_bar(sender, "Resampling Analysis", "Bootstrap Progress", bootstrap_n) + # else: + # progress = None + # f = 0.5 # count the observation as half less than itself f = 0 for i in range(bootstrap_n): @@ -332,6 +340,8 @@ def bootstrap_means(bootstrap_n, boot_data, obs_mean, pooled_var, random_effects # elif tmp_mean_e == mean_e: # f += 0.5 all_means.append(tmp_mean) + if progress_bar is not None: + progress_bar.setValue(progress_bar.value() + 1) all_means.sort() lower_index = round((bootstrap_n + 1) * alpha / 2) upper_index = round(bootstrap_n - (bootstrap_n + 1) * alpha / 2) @@ -558,7 +568,8 @@ def output_filtered_bad(filtered: list, bad_data: list) -> list: # ---------- basic meta-analysis ---------- -def simple_meta_analysis(data, options, decimal_places: int = 4, alpha: float = 0.05, norm_ci: bool = True): +def simple_meta_analysis(data, options, decimal_places: int = 4, alpha: float = 0.05, norm_ci: bool = True, + sender=None): # filter and prepare data for analysis effect_sizes = options.effect_data variances = options.effect_vars @@ -616,9 +627,16 @@ def simple_meta_analysis(data, options, decimal_places: int = 4, alpha: float = else: lower_ci, upper_ci = scipy.stats.t.interval(alpha=1-alpha, df=df, loc=mean_e, scale=math.sqrt(var_e)) + if (options.bootstrap_mean is not None) and (sender is not None): + progress_bar = MetaWinWidgets.progress_bar(sender, get_text("Resampling Progress"), + get_text("Conducting Bootstrap Analysis"), + options.bootstrap_mean) + else: + progress_bar = None lower_bs_ci, upper_bs_ci, lower_bias_ci, upper_bias_ci = bootstrap_means(options.bootstrap_mean, boot_data, mean_e, pooled_var, - options.random_effects, alpha) + options.random_effects, alpha, + progress_bar=progress_bar) plot_order = 0 mean_data = mean_data_tuple(get_text("Mean"), plot_order, n, mean_e, median_e, var_e, mean_v, @@ -684,7 +702,8 @@ def check_data_for_group(output_blocks, n, group_cnts, group_label) -> bool: return all_good -def grouped_meta_analysis(data, options, decimal_places: int = 4, alpha: float = 0.05, norm_ci: bool = True): +def grouped_meta_analysis(data, options, decimal_places: int = 4, alpha: float = 0.05, norm_ci: bool = True, + sender=None): # filter and prepare data for analysis effect_sizes = options.effect_data variances = options.effect_vars @@ -757,6 +776,19 @@ def grouped_meta_analysis(data, options, decimal_places: int = 4, alpha: float = qe = 0 chart_order = 2 group_i2_values = [] + + if ((options.bootstrap_mean is not None) or (options.randomize_model is not None)) and (sender is not None): + if options.randomize_model is not None: + total_steps = options.randomize_model + else: + total_steps = 0 + if options.bootstrap_mean is not None: + total_steps += (g_cnt + 1)*options.bootstrap_mean + progress_bar = MetaWinWidgets.progress_bar(sender, get_text("Resampling Progress"), + get_text("Conducting Bootstrap Analysis"), total_steps) + else: + progress_bar = None + for group in group_names: group_mask = [g == group for g in group_data] group_e = e_data[group_mask] @@ -777,7 +809,7 @@ def grouped_meta_analysis(data, options, decimal_places: int = 4, alpha: float = (group_lower_bs, group_upper_bs, group_lower_bias, group_upper_bias) = bootstrap_means(options.bootstrap_mean, group_boot, group_mean, pooled_var, options.random_effects, - alpha) + alpha, progress_bar=progress_bar) group_het_values.append(heterogeneity_test_tuple(group + " (within)", group_qw, group_df, group_p, "")) group_i2, group_i2_lower, group_i2_upper = calc_i2(group_qw, group_n, alpha) @@ -795,7 +827,8 @@ def grouped_meta_analysis(data, options, decimal_places: int = 4, alpha: float = lower_ci, upper_ci = scipy.stats.t.interval(alpha=1 - alpha, df=n-1, loc=mean_e, scale=math.sqrt(var_e)) lower_bs_ci, upper_bs_ci, lower_bias_ci, upper_bias_ci = bootstrap_means(options.bootstrap_mean, boot_data, mean_e, pooled_var, - options.random_effects, alpha) + options.random_effects, alpha, + progress_bar=progress_bar) global_mean_data = mean_data_tuple(get_text("Global"), 0, n, mean_e, median_e, var_e, mean_v, lower_ci, upper_ci, lower_bs_ci, upper_bs_ci, lower_bias_ci, upper_bias_ci) @@ -823,6 +856,8 @@ def grouped_meta_analysis(data, options, decimal_places: int = 4, alpha: float = rand_p_dec = max(decimal_places, math.ceil(math.log10(nreps+1))) cnt = 1 rng = numpy.random.default_rng() + if progress_bar is not None: + progress_bar.setLabelText(get_text("Conducting Randomization Analysis")) for rep in range(nreps): tmp_qe = 0 # rand_group_data = rng.permutation(group_data) @@ -836,6 +871,8 @@ def grouped_meta_analysis(data, options, decimal_places: int = 4, alpha: float = tmp_qmodel = qt - tmp_qe if tmp_qmodel >= qm: cnt += 1 + if progress_bar is not None: + progress_bar.setValue(progress_bar.value() + 1) p_random = cnt / (nreps + 1) p_random_str = format(p_random, inline_float(rand_p_dec)) else: @@ -885,7 +922,8 @@ def grouped_meta_analysis(data, options, decimal_places: int = 4, alpha: float = # ---------- cumulative meta-analysis ---------- -def cumulative_meta_analysis(data, options, decimal_places: int = 4, alpha: float = 0.05, norm_ci: bool = True): +def cumulative_meta_analysis(data, options, decimal_places: int = 4, alpha: float = 0.05, norm_ci: bool = True, + sender=None): # filter and prepare data for analysis effect_sizes = options.effect_data variances = options.effect_vars @@ -930,6 +968,13 @@ def cumulative_meta_analysis(data, options, decimal_places: int = 4, alpha: floa cumulative_means = [] cumulative_het = [] + if (options.bootstrap_mean is not None) and (sender is not None): + total_steps = (n - 1)*options.bootstrap_mean + progress_bar = MetaWinWidgets.progress_bar(sender, get_text("Resampling Progress"), + get_text("Conducting Bootstrap Analysis"), total_steps) + else: + progress_bar = None + chart_order = 0 for ns in range(2, n+1): ns_label = get_text("{} studies").format(ns) @@ -952,7 +997,8 @@ def cumulative_meta_analysis(data, options, decimal_places: int = 4, alpha: floa lower_ci, upper_ci = scipy.stats.t.interval(alpha=1-alpha, df=df, loc=mean_e, scale=math.sqrt(var_e)) lower_bs_ci, upper_bs_ci, lower_bias_ci, upper_bias_ci = bootstrap_means(options.bootstrap_mean, tmp_boot, mean_e, pooled_var, - options.random_effects, alpha) + options.random_effects, alpha, + progress_bar=progress_bar) mean_data = mean_data_tuple(ns_label, chart_order, ns, mean_e, median_e, var_e, mean_v, lower_ci, upper_ci, lower_bs_ci, upper_bs_ci, lower_bias_ci, upper_bias_ci) cumulative_means.append(mean_data) @@ -994,7 +1040,8 @@ def calculate_regression_ma_values(e_data, w_data, x_data, sum_w, sum_we, qt): return qm, qe, b1_slope, b0_intercept, var_b1, var_b0, sum_wx, sum_wx2 -def regression_meta_analysis(data, options, decimal_places: int = 4, alpha: float = 0.05, norm_ci: bool = True): +def regression_meta_analysis(data, options, decimal_places: int = 4, alpha: float = 0.05, norm_ci: bool = True, + sender=None): # filter and prepare data for analysis effect_sizes = options.effect_data variances = options.effect_vars @@ -1057,6 +1104,19 @@ def regression_meta_analysis(data, options, decimal_places: int = 4, alpha: floa sum_ws = sum_w sum_wse = sum_we + if ((options.bootstrap_mean is not None) or (options.randomize_model is not None)) and (sender is not None): + if options.randomize_model is not None: + total_steps = options.randomize_model + else: + total_steps = 0 + if options.bootstrap_mean is not None: + total_steps += options.bootstrap_mean + progress_bar = MetaWinWidgets.progress_bar(sender, get_text("Resampling Progress"), + get_text("Conducting Bootstrap Analysis"), + total_steps) + else: + progress_bar = None + median_e = median_effect(e_data, ws_data) mean_v = numpy.sum(v_data) / n if norm_ci: @@ -1065,7 +1125,8 @@ def regression_meta_analysis(data, options, decimal_places: int = 4, alpha: floa lower_ci, upper_ci = scipy.stats.t.interval(alpha=1 - alpha, df=n-1, loc=mean_e, scale=math.sqrt(var_e)) lower_bs_ci, upper_bs_ci, lower_bias_ci, upper_bias_ci = bootstrap_means(options.bootstrap_mean, boot_data, mean_e, pooled_var, - options.random_effects, alpha) + options.random_effects, alpha, + progress_bar=progress_bar) mean_data = mean_data_tuple(get_text("Global"), 0, n, mean_e, median_e, var_e, mean_v, lower_ci, upper_ci, lower_bs_ci, upper_bs_ci, lower_bias_ci, upper_bias_ci) @@ -1085,6 +1146,8 @@ def regression_meta_analysis(data, options, decimal_places: int = 4, alpha: floa # randomization test if options.randomize_model: + if progress_bar is not None: + progress_bar.setLabelText(get_text("Conducting Randomization Analysis")) nreps = options.randomize_model # decimal places to use for randomization-based p-value rand_p_dec = max(decimal_places, math.ceil(math.log10(nreps+1))) @@ -1096,6 +1159,8 @@ def regression_meta_analysis(data, options, decimal_places: int = 4, alpha: floa _, _, _, _) = calculate_regression_ma_values(rand_e_data, ws_data, x_data, sum_ws, sum_wse, qt) if rand_qm >= qm: cnt_q += 1 + if progress_bar is not None: + progress_bar.setValue(progress_bar.value() + 1) p_random = cnt_q / (nreps + 1) p_random_str = format(p_random, inline_float(rand_p_dec)) else: @@ -1185,7 +1250,8 @@ def calculate_glm(e: numpy.array, x: numpy.array, w: numpy.array): return qm, qe, beta, xtwxinv -def complex_meta_analysis(data, options, decimal_places: int = 4, alpha: float = 0.05, norm_ci: bool = True): +def complex_meta_analysis(data, options, decimal_places: int = 4, alpha: float = 0.05, norm_ci: bool = True, + sender=None): # filter and prepare data for analysis effect_sizes = options.effect_data variances = options.effect_vars @@ -1306,6 +1372,19 @@ def complex_meta_analysis(data, options, decimal_places: int = 4, alpha: float = pqe = 1 - scipy.stats.chi2.cdf(qe, df=dfe) pqm = 1 - scipy.stats.chi2.cdf(qm, df=dfm) + if ((options.bootstrap_mean is not None) or (options.randomize_model is not None)) and (sender is not None): + if options.randomize_model is not None: + total_steps = options.randomize_model + else: + total_steps = 0 + if options.bootstrap_mean is not None: + total_steps += options.bootstrap_mean + progress_bar = MetaWinWidgets.progress_bar(sender, get_text("Resampling Progress"), + get_text("Conducting Bootstrap Analysis"), + total_steps) + else: + progress_bar = None + # basic global calcs mean_e, var_e, _, _, _, _ = mean_effect_var_and_q(e_data, ws_data) mean_v = numpy.sum(v_data) / n @@ -1315,7 +1394,8 @@ def complex_meta_analysis(data, options, decimal_places: int = 4, alpha: float = lower_ci, upper_ci = scipy.stats.t.interval(alpha=1 - alpha, df=df, loc=mean_e, scale=math.sqrt(var_e)) lower_bs_ci, upper_bs_ci, lower_bias_ci, upper_bias_ci = bootstrap_means(options.bootstrap_mean, boot_data, mean_e, pooled_var, - options.random_effects, alpha) + options.random_effects, alpha, + progress_bar=progress_bar) mean_data = mean_data_tuple(get_text("Global"), 0, n, mean_e, median_e, var_e, mean_v, lower_ci, upper_ci, lower_bs_ci, upper_bs_ci, lower_bias_ci, upper_bias_ci) @@ -1328,6 +1408,8 @@ def complex_meta_analysis(data, options, decimal_places: int = 4, alpha: float = # randomization test if options.randomize_model: + if progress_bar is not None: + progress_bar.setLabelText(get_text("Conducting Randomization Analysis")) nreps = options.randomize_model # decimal places to use for randomization-based p-value rand_p_dec = max(decimal_places, math.ceil(math.log10(nreps+1))) @@ -1338,6 +1420,8 @@ def complex_meta_analysis(data, options, decimal_places: int = 4, alpha: float = rand_qm, _, _, _ = calculate_glm(rand_e_data, x_data, w_matrix) if rand_qm >= qm: cnt_q += 1 + if progress_bar is not None: + progress_bar.setValue(progress_bar.value() + 1) p_random = cnt_q / (nreps + 1) p_random_str = format(p_random, inline_float(rand_p_dec)) else: @@ -1438,8 +1522,17 @@ def qe(self, index): sn += cn return sq, sn + def nested_count(self): + """ + count the total number of nested groups, including this one + """ + count = 1 + for c in self.children: + count += c.nested_count() + return count + def group_calculations(self, e, w, chart_order: int, boot_data, bootstrap_mean, alpha: float = 0.05, - norm_ci: bool = True): + norm_ci: bool = True, progress_bar=None): chart_order += 1 mean_output = [] het_output = [] @@ -1461,7 +1554,7 @@ def group_calculations(self, e, w, chart_order: int, boot_data, bootstrap_mean, scale=math.sqrt(group_var)) (group_lower_bs, group_upper_bs, group_lower_bias, group_upper_bias) = bootstrap_means(bootstrap_mean, group_boot, self.mean, - 0, False, alpha) + 0, False, alpha, progress_bar=progress_bar) if self.index > 0: indent = " " + "→ "*self.index else: @@ -1472,7 +1565,8 @@ def group_calculations(self, e, w, chart_order: int, boot_data, bootstrap_mean, group_lower_bias, group_upper_bias)) for child in self.children: child_het, child_mean, chart_order = child.group_calculations(e, w, chart_order, boot_data, - bootstrap_mean, alpha) + bootstrap_mean, alpha, + progress_bar=progress_bar) het_output.extend(child_het) mean_output.extend(child_mean) if len(self.children) > 0: @@ -1526,7 +1620,8 @@ def find_next_nested_level(index, group_data, parent) -> list: return group_list -def nested_meta_analysis(data, options, decimal_places: int = 4, alpha: float = 0.05, norm_ci: bool = True): +def nested_meta_analysis(data, options, decimal_places: int = 4, alpha: float = 0.05, norm_ci: bool = True, + sender=None): # filter and prepare data for analysis effect_sizes = options.effect_data variances = options.effect_vars @@ -1566,7 +1661,6 @@ def nested_meta_analysis(data, options, decimal_places: int = 4, alpha: float = boot_data = numpy.array(boot_data) top_level = find_next_nested_level(0, group_data, None) - output_blocks = output_filtered_bad(filtered, bad_data) n = len(e_data) @@ -1582,6 +1676,23 @@ def nested_meta_analysis(data, options, decimal_places: int = 4, alpha: float = if check_data_for_nested(output_blocks, n, top_level, [level.label for level in group_levels]): output_blocks.append([get_text("{} studies will be included in this analysis").format(n)]) + total_groups = 0 + for group in top_level: + total_groups += group.nested_count() + + if ((options.bootstrap_mean is not None) or (options.randomize_model is not None)) and (sender is not None): + if options.randomize_model is not None: + total_steps = options.randomize_model + else: + total_steps = 0 + if options.bootstrap_mean is not None: + total_steps += options.bootstrap_mean * (total_groups + 1) + progress_bar = MetaWinWidgets.progress_bar(sender, get_text("Resampling Progress"), + get_text("Conducting Bootstrap Analysis"), + total_steps) + else: + progress_bar = None + mean_e, var_e, qt, sum_w, sum_w2, sum_ew = mean_effect_var_and_q(e_data, w_data) median_e = median_effect(e_data, w_data) mean_v = numpy.sum(v_data) / n @@ -1590,7 +1701,8 @@ def nested_meta_analysis(data, options, decimal_places: int = 4, alpha: float = else: lower_ci, upper_ci = scipy.stats.t.interval(alpha=1 - alpha, df=n-1, loc=mean_e, scale=math.sqrt(var_e)) lower_bs_ci, upper_bs_ci, lower_bias_ci, upper_bias_ci = bootstrap_means(options.bootstrap_mean, boot_data, - mean_e, 0, False, alpha) + mean_e, 0, False, alpha, + progress_bar=progress_bar) global_mean_data = mean_data_tuple(get_text("Global"), 0, n, mean_e, median_e, var_e, mean_v, lower_ci, upper_ci, lower_bs_ci, upper_bs_ci, lower_bias_ci, upper_bias_ci) df = n-1 @@ -1607,7 +1719,8 @@ def nested_meta_analysis(data, options, decimal_places: int = 4, alpha: float = group_mean_values = [] for group in top_level: het_out, mean_out, chart_order = group.group_calculations(e_data, w_data, chart_order, boot_data, - options.bootstrap_mean, alpha) + options.bootstrap_mean, alpha, + progress_bar=progress_bar) group_het_values.extend(het_out) group_mean_values.extend(mean_out) @@ -1644,6 +1757,8 @@ def nested_meta_analysis(data, options, decimal_places: int = 4, alpha: float = # randomization test if options.randomize_model: + if progress_bar is not None: + progress_bar.setLabelText(get_text("Conducting Randomization Analysis")) nreps = options.randomize_model # decimal places to use for randomization-based p-value rand_p_dec = max(decimal_places, math.ceil(math.log10(nreps+1))) @@ -1662,6 +1777,8 @@ def nested_meta_analysis(data, options, decimal_places: int = 4, alpha: float = tmp_qm += tq if tmp_qm >= model_het_values[i].q: cnt_list[i] += 1 + if progress_bar is not None: + progress_bar.setValue(progress_bar.value() + 1) for i in range(len(group_levels)): p_random = cnt_list[i] / (nreps + 1) p_random_str = format(p_random, inline_float(rand_p_dec)) @@ -1886,7 +2003,8 @@ def phylogenetic_correlation(tip_names, root): return p -def phylogenetic_meta_analysis(data, options, tree, decimal_places: int = 4, alpha: float = 0.05, norm_ci: bool = True): +def phylogenetic_meta_analysis(data, options, tree, decimal_places: int = 4, alpha: float = 0.05, norm_ci: bool = True, + sender=None): # filter and prepare data for analysis effect_sizes = options.effect_data variances = options.effect_vars @@ -2124,7 +2242,8 @@ def phylogenetic_meta_analysis(data, options, tree, decimal_places: int = 4, alp # ---------- jackknife meta-analysis ---------- -def jackknife_meta_analysis(data, options, decimal_places: int = 4, alpha: float = 0.05, norm_ci: bool = True): +def jackknife_meta_analysis(data, options, decimal_places: int = 4, alpha: float = 0.05, norm_ci: bool = True, + sender=None): # filter and prepare data for analysis effect_sizes = options.effect_data variances = options.effect_vars @@ -2163,6 +2282,14 @@ def jackknife_meta_analysis(data, options, decimal_places: int = 4, alpha: float if n > 1: output_blocks.append([get_text("{} studies will be included in this analysis").format(n)]) + if (options.bootstrap_mean is not None) and (sender is not None): + total_steps = options.bootstrap_mean*(n + 1) + progress_bar = MetaWinWidgets.progress_bar(sender, get_text("Resampling Progress"), + get_text("Conducting Bootstrap Analysis"), + total_steps) + else: + progress_bar = None + # global value df = n - 1 mean_e, var_e, qt, sum_w, sum_w2, sum_ew = mean_effect_var_and_q(e_data, w_data) @@ -2184,7 +2311,8 @@ def jackknife_meta_analysis(data, options, decimal_places: int = 4, alpha: float lower_bs_ci, upper_bs_ci, lower_bias_ci, upper_bias_ci = bootstrap_means(options.bootstrap_mean, boot_data, mean_e, pooled_var, - options.random_effects, alpha) + options.random_effects, alpha, + progress_bar=progress_bar) plot_order = 0 mean_data = mean_data_tuple(get_text("Mean"), plot_order, n, mean_e, median_e, var_e, mean_v, lower_ci, upper_ci, lower_bs_ci, upper_bs_ci, lower_bias_ci, upper_bias_ci) @@ -2226,7 +2354,8 @@ def jackknife_meta_analysis(data, options, decimal_places: int = 4, alpha: float lower_ci, upper_ci = scipy.stats.t.interval(alpha=1-alpha, df=df, loc=mean_e, scale=math.sqrt(var_e)) lower_bs_ci, upper_bs_ci, lower_bias_ci, upper_bias_ci = bootstrap_means(options.bootstrap_mean, tmp_boot, mean_e, pooled_var, - options.random_effects, alpha) + options.random_effects, alpha, + progress_bar=progress_bar) mean_data = mean_data_tuple(j_label, plot_order, n-1, mean_e, median_e, var_e, mean_v, lower_ci, upper_ci, lower_bs_ci, upper_bs_ci, lower_bias_ci, upper_bias_ci) jackknife_means.append(mean_data) @@ -2309,7 +2438,7 @@ def kendalls_tau(e_ranks, x_ranks): return tau -def rank_correlation_analysis(data, options, decimal_places: int = 4): +def rank_correlation_analysis(data, options, decimal_places: int = 4, sender=None): # filter and prepare data for analysis effect_sizes = options.effect_data variances = options.effect_vars @@ -2387,6 +2516,13 @@ def rank_correlation_analysis(data, options, decimal_places: int = 4): # test with randomization nreps = options.randomize_model + + if sender is not None: + progress_bar = MetaWinWidgets.progress_bar(sender, get_text("Resampling Progress"), + get_text("Conducting Randomization Analysis"), nreps) + else: + progress_bar = None + # decimal places to use for randomization-based p-value rand_p_dec = max(decimal_places, math.ceil(math.log10(nreps+1))) cnt_r = 1 @@ -2399,6 +2535,9 @@ def rank_correlation_analysis(data, options, decimal_places: int = 4): rand_r = correlation(e_ranks, x_ranks) if abs(rand_r) >= abs(r): cnt_r += 1 + if progress_bar is not None: + progress_bar.setValue(progress_bar.value() + 1) + p_random = cnt_r / (nreps + 1) p_random_str = format(p_random, inline_float(rand_p_dec)) rstr = format(r, inline_float(decimal_places)) diff --git a/src/MetaWinLanguage.py b/src/MetaWinLanguage.py index 064470f..9b53e39 100644 --- a/src/MetaWinLanguage.py +++ b/src/MetaWinLanguage.py @@ -47,6 +47,8 @@ "Complex/GLM": "Complex/GLM", "Complex/GLM Meta-Analysis": "Complex/GLM Meta-Analysis", "Compute": "Compute", + "Conducting Bootstrap Analysis": "Conducting Bootstrap Analysis", + "Conducting Randomization Analysis": "Conducting Randomization Analysis", "Confidence Intervals": "Confidence Intervals", "Continuous Variables": "Continuous Variables", "Continuous Independent Variables(s)": "Continuous Independent Variable(s)", @@ -299,6 +301,7 @@ "Rank Correlation Analysis": "Rank Correlation Analysis", "Rank Correlation Method": "Rank Correlation Method", "Rank Correlation Results": "Rank Correlation Results", + "Resampling Progress": "Resampling Progress", "Correlate Effect Size with": "Correlate Effect Size with", "References": "References", "Regression Results": "Regression Results", diff --git a/src/MetaWinWidgets.py b/src/MetaWinWidgets.py index a1c7ddb..b1f240a 100644 --- a/src/MetaWinWidgets.py +++ b/src/MetaWinWidgets.py @@ -5,7 +5,7 @@ from typing import Optional from PyQt6.QtWidgets import QPushButton, QVBoxLayout, QHBoxLayout, QListWidget, QAbstractItemView, QListWidgetItem, \ - QLabel, QComboBox, QCheckBox, QGroupBox, QGridLayout, QLineEdit, QColorDialog + QLabel, QComboBox, QCheckBox, QGroupBox, QGridLayout, QLineEdit, QColorDialog, QProgressDialog from PyQt6.QtGui import QIcon, QColor, QDoubleValidator from PyQt6 import QtCore @@ -250,3 +250,15 @@ def add_chart_line_edits(color_text, color, width_text, width, style_text, style width_label, width_box = add_chart_line_width(width_text, width) style_label, style_box = add_chart_line_style(style_text, style, line_styles) return color_label, color_button, width_label, width_box, style_label, style_box + + +def progress_bar(sender, title: str, text: str, max_val: int): + pb = QProgressDialog(sender) + pb.setWindowModality(QtCore.Qt.WindowModality.WindowModal) + pb.setWindowTitle(title) + pb.setLabelText(text) + pb.setMinimum(0) + pb.setMaximum(max_val) + pb.setCancelButton(None) + pb.setValue(0) + return pb