gwaslab 3.4.38__py3-none-any.whl → 3.4.39__py3-none-any.whl

gwaslab/ldsc_regressions.py

@@ -0,0 +1,747 @@
+'''
+(c) 2014 Brendan Bulik-Sullivan and Hilary Finucane
+
+Estimators of heritability and genetic correlation.
+
+Shape convention is (n_snp, n_annot) for all classes.
+Last column = intercept.
+
+'''
+from __future__ import division
+import numpy as np
+import pandas as pd
+from scipy.stats import norm, chi2
+import gwaslab.ldsc_jackknife as jk
+from gwaslab.ldsc_irwls import IRWLS
+from scipy.stats import t as tdist
+from collections import namedtuple
+np.seterr(divide='raise', invalid='raise')
+
+log_prefix = '                       -'
+log_prefix_short = '  -'
+s = lambda x: remove_brackets(str(np.matrix(x)))
+
+def xrange(*args):
+    return range(*args)
+
+def update_separators(s, ii):
+    '''s are separators with ii masked. Returns unmasked separators.'''
+    maplist = np.arange(len(ii))[np.squeeze(ii)]
+    mask_to_unmask = lambda i: maplist[i]
+    t = np.apply_along_axis(mask_to_unmask, 0, s[1:-1])
+    t = np.hstack(((0), t, (len(ii))))
+    return t
+
+
+def p_z_norm(est, se):
+    '''Convert estimate and se to Z-score and P-value.'''
+    try:
+        Z = est / se
+    except (FloatingPointError, ZeroDivisionError):
+        Z = float('inf')
+
+    P = chi2.sf(Z ** 2, 1, loc=0, scale=1)  # 0 if Z=inf
+    return P, Z
+
+
+def remove_brackets(x):
+    '''Get rid of brackets and trailing whitespace in numpy arrays.'''
+    return x.replace('[', '').replace(']', '').strip()
+
+
+def append_intercept(x):
+    '''
+    Appends an intercept term to the design matrix for a linear regression.
+
+    Parameters
+    ----------
+    x : np.matrix with shape (n_row, n_col)
+        Design matrix. Columns are predictors; rows are observations.
+
+    Returns
+    -------
+    x_new : np.matrix with shape (n_row, n_col+1)
+        Design matrix with intercept term appended.
+
+    '''
+    n_row = x.shape[0]
+    intercept = np.ones((n_row, 1))
+    x_new = np.concatenate((x, intercept), axis=1)
+    return x_new
+
+
+def remove_intercept(x):
+    '''Removes the last column.'''
+    n_col = x.shape[1]
+    return x[:, 0:n_col - 1]
+
+
+def gencov_obs_to_liab(gencov_obs, P1, P2, K1, K2):
+    '''
+    Converts genetic covariance on the observed scale in an ascertained sample to genetic
+    covariance on the liability scale in the population
+
+    Parameters
+    ----------
+    gencov_obs : float
+        Genetic covariance on the observed scale in an ascertained sample.
+    P1, P2 : float in (0,1)
+        Prevalences of phenotypes 1,2 in the sample.
+    K1, K2 : float in (0,1)
+        Prevalences of phenotypes 1,2 in the population.
+
+    Returns
+    -------
+    gencov_liab : float
+        Genetic covariance between liabilities in the population.
+
+    Note: if a trait is a QT, set P = K = None.
+
+    '''
+    c1 = 1
+    c2 = 1
+    if P1 is not None and K1 is not None:
+        c1 = np.sqrt(h2_obs_to_liab(1, P1, K1))
+    if P2 is not None and K2 is not None:
+        c2 = np.sqrt(h2_obs_to_liab(1, P2, K2))
+
+    return gencov_obs * c1 * c2
+
+
+def h2_obs_to_liab(h2_obs, P, K):
+    '''
+    Converts heritability on the observed scale in an ascertained sample to heritability
+    on the liability scale in the population.
+
+    Parameters
+    ----------
+    h2_obs : float
+        Heritability on the observed scale in an ascertained sample.
+    P : float in (0,1)
+        Prevalence of the phenotype in the sample.
+    K : float in (0,1)
+        Prevalence of the phenotype in the population.
+
+    Returns
+    -------
+    h2_liab : float
+        Heritability of liability in the population.
+
+    '''
+    if np.isnan(P) and np.isnan(K):
+        return h2_obs
+    if K <= 0 or K >= 1:
+        raise ValueError('K must be in the range (0,1)')
+    if P <= 0 or P >= 1:
+        raise ValueError('P must be in the range (0,1)')
+
+    thresh = norm.isf(K)
+    conversion_factor = K ** 2 * \
+        (1 - K) ** 2 / (P * (1 - P) * norm.pdf(thresh) ** 2)
+    return h2_obs * conversion_factor
+
+
+class LD_Score_Regression(object):
+
+    def __init__(self, y, x, w, N, M, n_blocks, intercept=None, slow=False, step1_ii=None, old_weights=False):
+        for i in [y, x, w, M, N]:
+            try:
+                if len(i.shape) != 2:
+                    raise TypeError('Arguments must be 2D arrays.')
+            except AttributeError:
+                raise TypeError('Arguments must be arrays.')
+
+        n_snp, self.n_annot = x.shape
+        if any(i.shape != (n_snp, 1) for i in [y, w, N]):
+            raise ValueError(
+                'N, weights and response (z1z2 or chisq) must have shape (n_snp, 1).')
+        if M.shape != (1, self.n_annot):
+            raise ValueError('M must have shape (1, n_annot).')
+
+        M_tot = float(np.sum(M))
+        x_tot = np.sum(x, axis=1).reshape((n_snp, 1))
+        self.constrain_intercept = intercept is not None
+        self.intercept = intercept
+        self.n_blocks = n_blocks
+        tot_agg = self.aggregate(y, x_tot, N, M_tot, intercept)
+        initial_w = self._update_weights(
+            x_tot, w, N, M_tot, tot_agg, intercept)
+        Nbar = np.mean(N)  # keep condition number low
+        x = np.multiply(N, x) / Nbar
+        if not self.constrain_intercept:
+            x, x_tot = append_intercept(x), append_intercept(x_tot)
+            yp = y
+        else:
+            yp = y - intercept
+            self.intercept_se = 'NA'
+        del y
+        self.twostep_filtered = None
+        if step1_ii is not None and self.constrain_intercept:
+            raise ValueError(
+                'twostep is not compatible with constrain_intercept.')
+        elif step1_ii is not None and self.n_annot > 1:
+            raise ValueError(
+                'twostep not compatible with partitioned LD Score yet.')
+        elif step1_ii is not None:
+            n1 = np.sum(step1_ii)
+            self.twostep_filtered = n_snp - n1
+            x1 = x[np.squeeze(step1_ii), :]
+            yp1, w1, N1, initial_w1 = map(
+                lambda a: a[step1_ii].reshape((n1, 1)), (yp, w, N, initial_w))
+            update_func1 = lambda a: self._update_func(
+                a, x1, w1, N1, M_tot, Nbar, ii=step1_ii)
+            step1_jknife = IRWLS(
+                x1, yp1, update_func1, n_blocks, slow=slow, w=initial_w1)
+            step1_int, _ = self._intercept(step1_jknife)
+            yp = yp - step1_int
+            x = remove_intercept(x)
+            x_tot = remove_intercept(x_tot)
+            update_func2 = lambda a: self._update_func(
+                a, x_tot, w, N, M_tot, Nbar, step1_int)
+            s = update_separators(step1_jknife.separators, step1_ii)
+            step2_jknife = IRWLS(
+                x, yp, update_func2, n_blocks, slow=slow, w=initial_w, separators=s)
+            c = np.sum(np.multiply(initial_w, x)) / \
+                np.sum(np.multiply(initial_w, np.square(x)))
+            jknife = self._combine_twostep_jknives(
+                step1_jknife, step2_jknife, M_tot, c, Nbar)
+        elif old_weights:
+            initial_w = np.sqrt(initial_w)
+            x = IRWLS._weight(x, initial_w)
+            y = IRWLS._weight(yp, initial_w)
+            jknife = jk.LstsqJackknifeFast(x, y, n_blocks)
+        else:
+            update_func = lambda a: self._update_func(
+                a, x_tot, w, N, M_tot, Nbar, intercept)
+            jknife = IRWLS(
+                x, yp, update_func, n_blocks, slow=slow, w=initial_w)
+
+        self.coef, self.coef_cov, self.coef_se = self._coef(jknife, Nbar)
+        self.cat, self.cat_cov, self.cat_se =\
+            self._cat(jknife, M, Nbar, self.coef, self.coef_cov)
+
+        self.tot, self.tot_cov, self.tot_se = self._tot(self.cat, self.cat_cov)
+        self.prop, self.prop_cov, self.prop_se =\
+            self._prop(jknife, M, Nbar, self.cat, self.tot)
+
+        self.enrichment, self.M_prop = self._enrichment(
+            M, M_tot, self.cat, self.tot)
+        if not self.constrain_intercept:
+            self.intercept, self.intercept_se = self._intercept(jknife)
+
+        self.jknife = jknife
+        self.tot_delete_values = self._delete_vals_tot(jknife, Nbar, M)
+        self.part_delete_values = self._delete_vals_part(jknife, Nbar, M)
+        if not self.constrain_intercept:
+            self.intercept_delete_values = jknife.delete_values[
+                :, self.n_annot]
+
+        self.M = M
+
+    @classmethod
+    def aggregate(cls, y, x, N, M, intercept=None):
+        if intercept is None:
+            intercept = cls.__null_intercept__
+
+        num = M * (np.mean(y) - intercept)
+        denom = np.mean(np.multiply(x, N))
+        return num / denom
+
+    def _update_func(self, x, ref_ld_tot, w_ld, N, M, Nbar, intercept=None, ii=None):
+        raise NotImplementedError
+
+    def _delete_vals_tot(self, jknife, Nbar, M):
+        '''Get delete values for total h2 or gencov.'''
+        n_annot = self.n_annot
+        tot_delete_vals = jknife.delete_values[
+            :, 0:n_annot]  # shape (n_blocks, n_annot)
+        # shape (n_blocks, 1)
+        tot_delete_vals = np.dot(tot_delete_vals, M.T) / Nbar
+        return tot_delete_vals
+
+    def _delete_vals_part(self, jknife, Nbar, M):
+        '''Get delete values for partitioned h2 or gencov.'''
+        n_annot = self.n_annot
+        return jknife.delete_values[:, 0:n_annot] / Nbar
+
+    def _coef(self, jknife, Nbar):
+        '''Get coefficient estimates + cov from the jackknife.'''
+        n_annot = self.n_annot
+        coef = jknife.est[0, 0:n_annot] / Nbar
+        coef_cov = jknife.jknife_cov[0:n_annot, 0:n_annot] / Nbar ** 2
+        coef_se = np.sqrt(np.diag(coef_cov))
+        return coef, coef_cov, coef_se
+
+    def _cat(self, jknife, M, Nbar, coef, coef_cov):
+        '''Convert coefficients to per-category h2 or gencov.'''
+        cat = np.multiply(M, coef)
+        cat_cov = np.multiply(np.dot(M.T, M), coef_cov)
+        cat_se = np.sqrt(np.diag(cat_cov))
+        return cat, cat_cov, cat_se
+
+    def _tot(self, cat, cat_cov):
+        '''Convert per-category h2 to total h2 or gencov.'''
+        tot = np.sum(cat)
+        tot_cov = np.sum(cat_cov)
+        tot_se = np.sqrt(tot_cov)
+        return tot, tot_cov, tot_se
+
+    def _prop(self, jknife, M, Nbar, cat, tot):
+        '''Convert total h2 and per-category h2 to per-category proportion h2 or gencov.'''
+        n_annot = self.n_annot
+        n_blocks = jknife.delete_values.shape[0]
+        numer_delete_vals = np.multiply(
+            M, jknife.delete_values[:, 0:n_annot]) / Nbar  # (n_blocks, n_annot)
+        denom_delete_vals = np.sum(
+            numer_delete_vals, axis=1).reshape((n_blocks, 1))
+        denom_delete_vals = np.dot(denom_delete_vals, np.ones((1, n_annot)))
+        prop = jk.RatioJackknife(
+            cat / tot, numer_delete_vals, denom_delete_vals)
+        return prop.est, prop.jknife_cov, prop.jknife_se
+
+    def _enrichment(self, M, M_tot, cat, tot):
+        '''Compute proportion of SNPs per-category enrichment for h2 or gencov.'''
+        M_prop = M / M_tot
+        enrichment = np.divide(cat, M) / (tot / M_tot)
+        return enrichment, M_prop
+
+    def _intercept(self, jknife):
+        '''Extract intercept and intercept SE from block jackknife.'''
+        n_annot = self.n_annot
+        intercept = jknife.est[0, n_annot]
+        intercept_se = jknife.jknife_se[0, n_annot]
+        return intercept, intercept_se
+
+    def _combine_twostep_jknives(self, step1_jknife, step2_jknife, M_tot, c, Nbar=1):
+        '''Combine free intercept and constrained intercept jackknives for --two-step.'''
+        n_blocks, n_annot = step1_jknife.delete_values.shape
+        n_annot -= 1
+        if n_annot > 2:
+            raise ValueError(
+                'twostep not yet implemented for partitioned LD Score.')
+
+        step1_int, _ = self._intercept(step1_jknife)
+        est = np.hstack(
+            (step2_jknife.est, np.array(step1_int).reshape((1, 1))))
+        delete_values = np.zeros((n_blocks, n_annot + 1))
+        delete_values[:, n_annot] = step1_jknife.delete_values[:, n_annot]
+        delete_values[:, 0:n_annot] = step2_jknife.delete_values -\
+            c * (step1_jknife.delete_values[:, n_annot] -
+                 step1_int).reshape((n_blocks, n_annot))  # check this
+        pseudovalues = jk.Jackknife.delete_values_to_pseudovalues(
+            delete_values, est)
+        jknife_est, jknife_var, jknife_se, jknife_cov = jk.Jackknife.jknife(
+            pseudovalues)
+        jknife = namedtuple('jknife',
+                            ['est', 'jknife_se', 'jknife_est', 'jknife_var', 'jknife_cov', 'delete_values'])
+        return jknife(est, jknife_se, jknife_est, jknife_var, jknife_cov, delete_values)
+
+
+class Hsq(LD_Score_Regression):
+
+    __null_intercept__ = 1
+
+    def __init__(self, y, x, w, N, M, n_blocks=200, intercept=None, slow=False, twostep=None, old_weights=False):
+        step1_ii = None
+        if twostep is not None:
+            step1_ii = y < twostep
+
+        LD_Score_Regression.__init__(self, y, x, w, N, M, n_blocks, intercept=intercept,
+                                     slow=slow, step1_ii=step1_ii, old_weights=old_weights)
+        self.mean_chisq, self.lambda_gc = self._summarize_chisq(y)
+        if not self.constrain_intercept:
+            self.ratio, self.ratio_se = self._ratio(
+                self.intercept, self.intercept_se, self.mean_chisq)
+
+    def _update_func(self, x, ref_ld_tot, w_ld, N, M, Nbar, intercept=None, ii=None):
+        '''
+        Update function for IRWLS
+
+        x is the output of np.linalg.lstsq.
+        x[0] is the regression coefficients
+        x[0].shape is (# of dimensions, 1)
+        the last element of x[0] is the intercept.
+
+        intercept is None --> free intercept
+        intercept is not None --> constrained intercept
+        '''
+        hsq = M * x[0][0] / Nbar
+        if intercept is None:
+            intercept = max(x[0][1])  # divide by zero error if intercept < 0
+        else:
+            if ref_ld_tot.shape[1] > 1:
+                raise ValueError(
+                    'Design matrix has intercept column for constrained intercept regression!')
+
+        ld = ref_ld_tot[:, 0].reshape(w_ld.shape)  # remove intercept
+        w = self.weights(ld, w_ld, N, M, hsq, intercept, ii)
+        return w
+
+    def _summarize_chisq(self, chisq):
+        '''Compute mean chi^2 and lambda_GC.'''
+        mean_chisq = np.mean(chisq)
+        # median and matrix don't play nice
+        lambda_gc = np.median(np.asarray(chisq)) / 0.4549
+        return mean_chisq, lambda_gc
+
+    def _ratio(self, intercept, intercept_se, mean_chisq):
+        '''Compute ratio (intercept - 1) / (mean chi^2 -1 ).'''
+        if mean_chisq > 1:
+            ratio_se = intercept_se / (mean_chisq - 1)
+            ratio = (intercept - 1) / (mean_chisq - 1)
+        else:
+            ratio = 'NA'
+            ratio_se = 'NA'
+
+        return ratio, ratio_se
+
+    def _overlap_output(self, category_names, overlap_matrix, M_annot, M_tot, print_coefficients):
+        '''LD Score regression summary for overlapping categories.'''
+        overlap_matrix_prop = np.zeros([self.n_annot,self.n_annot])
+        for i in range(self.n_annot):
+            overlap_matrix_prop[i, :] = overlap_matrix[i, :] / M_annot
+
+        prop_hsq_overlap = np.dot(
+            overlap_matrix_prop, self.prop.T).reshape((1, self.n_annot))
+        prop_hsq_overlap_var = np.diag(
+            np.dot(np.dot(overlap_matrix_prop, self.prop_cov), overlap_matrix_prop.T))
+        prop_hsq_overlap_se = np.sqrt(
+            np.maximum(0, prop_hsq_overlap_var)).reshape((1, self.n_annot))
+        one_d_convert = lambda x: np.array(x).reshape(np.prod(x.shape))
+        prop_M_overlap = M_annot / M_tot
+        enrichment = prop_hsq_overlap / prop_M_overlap
+        enrichment_se = prop_hsq_overlap_se / prop_M_overlap
+        overlap_matrix_diff = np.zeros([self.n_annot,self.n_annot])
+        for i in range(self.n_annot):
+            if not M_tot == M_annot[0,i]:
+                overlap_matrix_diff[i, :] = overlap_matrix[i,:]/M_annot[0,i] - \
+                    (M_annot - overlap_matrix[i,:]) / (M_tot-M_annot[0,i])
+
+        diff_est = np.dot(overlap_matrix_diff,self.coef)
+        diff_cov = np.dot(np.dot(overlap_matrix_diff,self.coef_cov),overlap_matrix_diff.T)
+        diff_se = np.sqrt(np.diag(diff_cov))
+        diff_p = ['NA' if diff_se[i]==0 else 2*tdist.sf(abs(diff_est[i]/diff_se[i]),self.n_blocks) \
+            for i in range(self.n_annot)]
+
+        df = pd.DataFrame({
+            'Category': category_names,
+            'Prop._SNPs': one_d_convert(prop_M_overlap),
+            'Prop._h2': one_d_convert(prop_hsq_overlap),
+            'Prop._h2_std_error': one_d_convert(prop_hsq_overlap_se),
+            'Enrichment': one_d_convert(enrichment),
+            'Enrichment_std_error': one_d_convert(enrichment_se),
+            'Enrichment_p':diff_p,
+            'Coefficient': one_d_convert(self.coef),
+            'Coefficient_std_error': self.coef_se,
+            'Coefficient_z-score': one_d_convert(self.coef) / one_d_convert(self.coef_se)
+        })
+        if print_coefficients:
+            df = df[['Category', 'Prop._SNPs', 'Prop._h2', 'Prop._h2_std_error',
+                    'Enrichment','Enrichment_std_error', 'Enrichment_p',
+                     'Coefficient', 'Coefficient_std_error','Coefficient_z-score']]
+        else:
+            df = df[['Category', 'Prop._SNPs', 'Prop._h2', 'Prop._h2_std_error',
+                    'Enrichment','Enrichment_std_error', 'Enrichment_p']]
+        return df
+
+
+    def summary(self, ref_ld_colnames=None, P=None, K=None, overlap=False):
+        '''Print summary of the LD Score Regression.'''
+        if P is not None and K is not None:
+            T = 'Liability'
+            c = h2_obs_to_liab(1, P, K)
+        else:
+            T = 'Observed'
+            c = 1
+
+        out = ['Total ' + T + ' scale h2: ' +
+               s(c * self.tot) + ' (' + s(c * self.tot_se) + ')']
+        if self.n_annot > 1:
+            if ref_ld_colnames is None:
+                ref_ld_colnames = ['CAT_' + str(i)
+                                   for i in xrange(self.n_annot)]
+
+            out.append(log_prefix_short+'Categories: ' + ' '.join(ref_ld_colnames))
+
+            if not overlap:
+                out.append(log_prefix+T + ' scale h2: ' + s(c * self.cat))
+                out.append(log_prefix+T + ' scale h2 SE: ' + s(c * self.cat_se))
+                out.append(log_prefix+'Proportion of SNPs: ' + s(self.M_prop))
+                out.append(log_prefix+'Proportion of h2g: ' + s(self.prop))
+                out.append(log_prefix+'Enrichment: ' + s(self.enrichment))
+                out.append(log_prefix+'Coefficients: ' + s(self.coef))
+                out.append(log_prefix+'Coefficient SE: ' + s(self.coef_se))
+
+        out.append(log_prefix+'Lambda GC: ' + s(self.lambda_gc))
+        out.append(log_prefix+'Mean Chi^2: ' + s(self.mean_chisq))
+        if self.constrain_intercept:
+            out.append(
+                log_prefix+'Intercept: constrained to {C}'.format(C=s(self.intercept)))
+        else:
+            out.append(
+                log_prefix+ 'Intercept: ' + s(self.intercept) + ' (' + s(self.intercept_se) + ')')
+            if self.mean_chisq > 1:
+                if self.ratio < 0:
+                    out.append(
+                      log_prefix+'Ratio < 0 (usually indicates GC correction).')
+                else:
+                    out.append(
+                      log_prefix+'Ratio: ' + s(self.ratio) + ' (' + s(self.ratio_se) + ')')
+            else:
+                out.append(log_prefix+'Ratio: NA (mean chi^2 < 1)')
+
+        return remove_brackets('\n'.join(out))
+
+    def _update_weights(self, ld, w_ld, N, M, hsq, intercept, ii=None):
+        if intercept is None:
+            intercept = self.__null_intercept__
+
+        return self.weights(ld, w_ld, N, M, hsq, intercept, ii)
+
+    @classmethod
+    def weights(cls, ld, w_ld, N, M, hsq, intercept=None, ii=None):
+        '''
+        Regression weights.
+
+        Parameters
+        ----------
+        ld : np.matrix with shape (n_snp, 1)
+            LD Scores (non-partitioned).
+        w_ld : np.matrix with shape (n_snp, 1)
+            LD Scores (non-partitioned) computed with sum r^2 taken over only those SNPs included
+            in the regression.
+        N :  np.matrix of ints > 0 with shape (n_snp, 1)
+            Number of individuals sampled for each SNP.
+        M : float > 0
+            Number of SNPs used for estimating LD Score (need not equal number of SNPs included in
+            the regression).
+        hsq : float in [0,1]
+            Heritability estimate.
+
+        Returns
+        -------
+        w : np.matrix with shape (n_snp, 1)
+            Regression weights. Approx equal to reciprocal of conditional variance function.
+
+        '''
+        M = float(M)
+        if intercept is None:
+            intercept = 1
+
+        hsq = max(hsq, 0.0)
+        hsq = min(hsq, 1.0)
+        ld = np.fmax(ld, 1.0)
+        w_ld = np.fmax(w_ld, 1.0)
+        c = hsq * N / M
+        het_w = 1.0 / (2 * np.square(intercept + np.multiply(c, ld)))
+        oc_w = 1.0 / w_ld
+        w = np.multiply(het_w, oc_w)
+        return w
+
+
+class Gencov(LD_Score_Regression):
+    __null_intercept__ = 0
+
+    def __init__(self, z1, z2, x, w, N1, N2, M, hsq1, hsq2, intercept_hsq1, intercept_hsq2,
+                 n_blocks=200, intercept_gencov=None, slow=False, twostep=None):
+        self.intercept_hsq1 = intercept_hsq1
+        self.intercept_hsq2 = intercept_hsq2
+        self.hsq1 = hsq1
+        self.hsq2 = hsq2
+        self.N1 = N1
+        self.N2 = N2
+        y = z1 * z2
+        step1_ii = None
+        if twostep is not None:
+            step1_ii = np.logical_and(z1**2 < twostep, z2**2 < twostep)
+
+        LD_Score_Regression.__init__(self, y, x, w, np.sqrt(N1 * N2), M, n_blocks,
+                                     intercept=intercept_gencov, slow=slow, step1_ii=step1_ii)
+        self.p, self.z = p_z_norm(self.tot, self.tot_se)
+        self.mean_z1z2 = np.mean(np.multiply(z1, z2))
+
+    def summary(self, ref_ld_colnames, P=None, K=None):
+        '''Print summary of the LD Score regression.'''
+        out = []
+        if P is not None and K is not None and\
+                all((i is not None for i in P)) and all((i is not None for i in K)):
+            T = 'Liability'
+            c = gencov_obs_to_liab(1, P[0], P[1], K[0], K[1])
+        else:
+            T = 'Observed'
+            c = 1
+
+        out.append('Total ' + T + ' scale gencov: ' +
+                   s(c * self.tot) + ' (' + s(c * self.tot_se) + ')')
+        if self.n_annot > 1:
+            out.append(log_prefix+'Categories: ' + str(' '.join(ref_ld_colnames)))
+            out.append(log_prefix+T + ' scale gencov: ' + s(c * self.cat))
+            out.append(log_prefix+T + ' scale gencov SE: ' + s(c * self.cat_se))
+            out.append(log_prefix+'Proportion of SNPs: ' + s(self.M_prop))
+            out.append(log_prefix+'Proportion of gencov: ' + s(self.prop))
+            out.append(log_prefix+'Enrichment: ' + s(self.enrichment))
+
+        out.append(log_prefix+'Mean z1*z2: ' + s(self.mean_z1z2))
+        if self.constrain_intercept:
+            out.append(
+                log_prefix+'Intercept: constrained to {C}'.format(C=s(self.intercept)))
+        else:
+            out.append(
+                log_prefix+'Intercept: ' + s(self.intercept) + ' (' + s(self.intercept_se) + ')')
+
+        return remove_brackets('\n'.join(out))
+
+    def _update_func(self, x, ref_ld_tot, w_ld, N, M, Nbar, intercept=None, ii=None):
+        '''
+        Update function for IRWLS
+        x is the output of np.linalg.lstsq.
+        x[0] is the regression coefficients
+        x[0].shape is (# of dimensions, 1)
+        the last element of x[0] is the intercept.
+
+        '''
+        rho_g = M * x[0][0] / Nbar
+        if intercept is None:  # if the regression includes an intercept
+            intercept = x[0][1]
+
+        # remove intercept if we have one
+        ld = ref_ld_tot[:, 0].reshape(w_ld.shape)
+        if ii is not None:
+            N1 = self.N1[ii].reshape((w_ld.shape))
+            N2 = self.N2[ii].reshape((w_ld.shape))
+        else:
+            N1 = self.N1
+            N2 = self.N2
+
+        return self.weights(ld, w_ld, N1, N2, np.sum(M), self.hsq1, self.hsq2, rho_g,
+                         intercept, self.intercept_hsq1, self.intercept_hsq2, ii)
+
+    def _update_weights(self, ld, w_ld, sqrt_n1n2, M, rho_g, intercept, ii=None):
+        '''Weight function with the same signature for Hsq and Gencov.'''
+        w = self.weights(ld, w_ld, self.N1, self.N2, M, self.hsq1, self.hsq2, rho_g,
+                         intercept, self.intercept_hsq1, self.intercept_hsq2)
+        return w
+
+    @classmethod
+    def weights(cls, ld, w_ld, N1, N2, M, h1, h2, rho_g, intercept_gencov=None,
+                intercept_hsq1=None, intercept_hsq2=None, ii=None):
+        '''
+        Regression weights.
+
+        Parameters
+        ----------
+        ld : np.matrix with shape (n_snp, 1)
+            LD Scores (non-partitioned)
+        w_ld : np.matrix with shape (n_snp, 1)
+            LD Scores (non-partitioned) computed with sum r^2 taken over only those SNPs included
+            in the regression.
+        M : float > 0
+            Number of SNPs used for estimating LD Score (need not equal number of SNPs included in
+            the regression).
+        N1, N2 :  np.matrix of ints > 0 with shape (n_snp, 1)
+            Number of individuals sampled for each SNP for each study.
+        h1, h2 : float in [0,1]
+            Heritability estimates for each study.
+        rhog : float in [0,1]
+            Genetic covariance estimate.
+        intercept : float
+            Genetic covariance intercept, on the z1*z2 scale (so should be Ns*rho/sqrt(N1*N2)).
+
+        Returns
+        -------
+        w : np.matrix with shape (n_snp, 1)
+            Regression weights. Approx equal to reciprocal of conditional variance function.
+
+        '''
+        M = float(M)
+        if intercept_gencov is None:
+            intercept_gencov = 0
+        if intercept_hsq1 is None:
+            intercept_hsq1 = 1
+        if intercept_hsq2 is None:
+            intercept_hsq2 = 1
+
+        h1, h2 = max(h1, 0.0), max(h2, 0.0)
+        h1, h2 = min(h1, 1.0), min(h2, 1.0)
+        rho_g = min(rho_g, 1.0)
+        rho_g = max(rho_g, -1.0)
+        ld = np.fmax(ld, 1.0)
+        w_ld = np.fmax(w_ld, 1.0)
+        a = np.multiply(N1, h1 * ld) / M + intercept_hsq1
+        b = np.multiply(N2, h2 * ld) / M + intercept_hsq2
+        sqrt_n1n2 = np.sqrt(np.multiply(N1, N2))
+        c = np.multiply(sqrt_n1n2, rho_g * ld) / M + intercept_gencov
+        try:
+            het_w = 1.0 / (np.multiply(a, b) + np.square(c))
+        except FloatingPointError:  # bizarre error; should never happen
+            raise FloatingPointError('Why did you set hsq intercept <= 0?')
+
+        oc_w = 1.0 / w_ld
+        w = np.multiply(het_w, oc_w)
+        return w
+
+
+class RG(object):
+
+    def __init__(self, z1, z2, x, w, N1, N2, M, intercept_hsq1=None, intercept_hsq2=None,
+                 intercept_gencov=None, n_blocks=200, slow=False, twostep=None):
+        self.intercept_gencov = intercept_gencov
+        self._negative_hsq = None
+        n_snp, n_annot = x.shape
+        hsq1 = Hsq(np.square(z1), x, w, N1, M, n_blocks=n_blocks, intercept=intercept_hsq1,
+                   slow=slow, twostep=twostep)
+        hsq2 = Hsq(np.square(z2), x, w, N2, M, n_blocks=n_blocks, intercept=intercept_hsq2,
+                   slow=slow, twostep=twostep)
+        gencov = Gencov(z1, z2, x, w, N1, N2, M, hsq1.tot, hsq2.tot, hsq1.intercept,
+                        hsq2.intercept, n_blocks, intercept_gencov=intercept_gencov, slow=slow,
+                        twostep=twostep)
+        gencov.N1 = None  # save memory
+        gencov.N2 = None
+        self.hsq1, self.hsq2, self.gencov = hsq1, hsq2, gencov
+        if (hsq1.tot <= 0 or hsq2.tot <= 0):
+            self._negative_hsq = True
+            self.rg_ratio = self.rg = self.rg_se = 'NA'
+            self.p = self.z = 'NA'
+        else:
+            rg_ratio = np.array(
+                gencov.tot / np.sqrt(hsq1.tot * hsq2.tot)).reshape((1, 1))
+            denom_delete_values = np.sqrt(
+                np.multiply(hsq1.tot_delete_values, hsq2.tot_delete_values))
+            rg = jk.RatioJackknife(
+                rg_ratio, gencov.tot_delete_values, denom_delete_values)
+            self.rg_jknife = float(rg.jknife_est)
+            self.rg_se = float(rg.jknife_se)
+            self.rg_ratio = float(rg_ratio)
+            self.p, self.z = p_z_norm(self.rg_ratio, self.rg_se)
+
+    def summary(self, silly=False):
+        '''Print output of Gencor object.'''
+        out = []
+        if self._negative_hsq:
+            out.append(log_prefix_short+'Genetic Correlation: nan (nan) (h2  out of bounds) ')
+            out.append(log_prefix+'Z-score: nan (nan) (h2  out of bounds)')
+            out.append(log_prefix+'P: nan (nan) (h2  out of bounds)')
+            out.append(log_prefix+'WARNING: One of the h2\'s was out of bounds.')
+            out.append(
+                log_prefix+'This usually indicates a data-munging error ' +
+                'or that h2 or N is low.')
+        elif (self.rg_ratio > 1.2 or self.rg_ratio < -1.2) and not silly:
+            out.append(log_prefix_short+'Genetic Correlation: nan (nan) (rg out of bounds) ')
+            out.append(log_prefix+'Z-score: nan (nan) (rg out of bounds)')
+            out.append(log_prefix+'P: nan (nan) (rg out of bounds)')
+            out.append(log_prefix+'WARNING: rg was out of bounds.')
+            if self.intercept_gencov is None:
+                out.append(
+                    log_prefix+'This often means that h2 is not significantly ' +
+                    'different from zero.')
+            else:
+                out.append(
+                           log_prefix+'This often means that you have constrained' +
+                           ' the intercepts to the wrong values.')
+        else:
+            out.append(
+                'Genetic Correlation: ' + s(self.rg_ratio) +
+                ' (' + s(self.rg_se) + ')')
+            out.append(log_prefix+'Z-score: ' + s(self.z))
+            out.append(log_prefix+'P: ' + s(self.p))
+        return remove_brackets('\n'.join(out))