Fixed database issues

README.md

@@ -1,22 +1,16 @@
 # SQUAAD ANALYSIS FRAMEWORK
 
-## Dependencies
-
-Install R on Ubuntu:
-
-`sudo apt-get install r-base`
-
 ## Installation
 
 `pip install squaad`
 
 ### Releases
 
-* V1.0 `https://github.com/fostiropoulos/squaad/releases/download/v1.0/squaad-1.0.tar.gz`
+* V2.0 `https://github.com/fostiropoulos/squaad/releases/download/v2.0/squaad-2.0.tar.gz`
 
 
 ### Install from Binary
-`pip install squaad-1.0.tar.gz`
+`pip install squaad-2.0.tar.gz`
 
 ## Usage
 

requirements.txt

@@ -1,5 +1,5 @@
-rpy2
-psycopg2
+psycopg2-binary
+xlwt
 GitPython
 SQLAlchemy
 imbalanced-learn

setup.py

@@ -5,14 +5,18 @@
 
 setuptools.setup(
     name="squaad",
-    version="1.0",
+    version="2.0",
     author='Iordanis Fostiropoulos',
     author_email='danny.fostiropoulos@gmail.com',
     description='Helper functions for running queries, ml pipeline, statistical analysis on SQUAAD framework',
     long_description=long_description,
     long_description_content_type="text/markdown",
     url="http://github.com/fostiropoulos/squaad",
     packages=setuptools.find_packages(),
+    package_data={
+   'squaad': ['sql/*']
+
+   },
     classifiers=[
         "Programming Language :: Python :: 3",
         "License :: OSI Approved :: Mozilla Public License 2.0 (MPL 2.0)",

squaad/db.py

@@ -23,6 +23,7 @@ def __init__(self, config_file , cache_folder=None):
 			config_file (str): path to config file that contains the connection information. Fromat, `{"pgsql":{"host":"","user":"","passwd":"","db":""}`
 			cache_folder (:obj:`str`, optional): path to the cache folder. None for no cache.
 		"""
+
 		if(cache_folder!=None):
 			if(not os.path.isdir(cache_folder)):
 				raise Exception("Cache folder %s doesn't exist"%cache_folder)
@@ -33,6 +34,7 @@ def __init__(self, config_file , cache_folder=None):
 		except Exception as e:
 			print("Could not laod configuration file: "+config_file)
 			raise e
+
 		try:
 			conn = psycopg2.connect("dbname='%s' user='%s' host='%s' password='%s'"%(config["pgsql"]["db"],config["pgsql"]["user"],config["pgsql"]["host"],config["pgsql"]["passwd"],))
 		except Exception as e:

squaad/db_lite.py

@@ -62,7 +62,7 @@ def connect_sqlalchemy_db(db_name, user):
 	return engine, meta
 
 
-class db(object):
+class db_lite(object):
 
 	def __init__(self, db_name: str, user: str):
 		self.engine, self.meta = connect_sqlalchemy_db(db_name, user)
@@ -112,4 +112,3 @@ def pd_to_table(self, df, name, if_exists='replace'):
 
 	def table_to_pd(self, name, cols=None, index_col=None, parse_dates=None):
 		return pd.read_sql_table(name, con=self.engine, columns=cols, index_col=index_col, parse_dates=parse_dates)
-

squaad/stats.py

@@ -1,82 +1,6 @@
-from rpy2.robjects.packages import importr
-import rpy2.robjects as robjects
+from pingouin import pairwise_gameshowell
 
 class stats():
-	def __init__(self):
-		base = importr('base')
-
-		robjects.r('''
-				games.howell <- function(grp, obs) {
-
-				  #Create combinations
-				  combs <- combn(unique(grp), 2)
-
-				  # Statistics that will be used throughout the calculations:
-				  # n = sample size of each group
-				  # groups = number of groups in data
-				  # Mean = means of each group sample
-				  # std = variance of each group sample
-				  n <- tapply(obs, grp, length)
-				  groups <- length(tapply(obs, grp, length))
-				  Mean <- tapply(obs, grp, mean)
-				  std <- tapply(obs, grp, var)
-
-				  statistics <- lapply(1:ncol(combs), function(x) {
-
-					mean.diff <- Mean[combs[2,x]] - Mean[combs[1,x]]
-
-					#t-values
-					t <- abs(Mean[combs[1,x]] - Mean[combs[2,x]]) / sqrt((std[combs[1,x]] / n[combs[1,x]]) + (std[combs[2,x]] / n[combs[2,x]]))
-
-					# Degrees of Freedom
-					df <- (std[combs[1,x]] / n[combs[1,x]] + std[combs[2,x]] / n[combs[2,x]])^2 / # Numerator Degrees of Freedom
-					  ((std[combs[1,x]] / n[combs[1,x]])^2 / (n[combs[1,x]] - 1) + # Part 1 of Denominator Degrees of Freedom
-						 (std[combs[2,x]] / n[combs[2,x]])^2 / (n[combs[2,x]] - 1)) # Part 2 of Denominator Degrees of Freedom
-
-					#p-values
-					p <- ptukey(t * sqrt(2), groups, df, lower.tail = FALSE)
-
-					# Sigma standard error
-					se <- sqrt(0.5 * (std[combs[1,x]] / n[combs[1,x]] + std[combs[2,x]] / n[combs[2,x]]))
-
-					# Upper Confidence Limit
-					upper.conf <- lapply(1:ncol(combs), function(x) {
-					  mean.diff + qtukey(p = 0.95, nmeans = groups, df = df) * se
-					})[[1]]
-
-					# Lower Confidence Limit
-					lower.conf <- lapply(1:ncol(combs), function(x) {
-					  mean.diff - qtukey(p = 0.95, nmeans = groups, df = df) * se
-					})[[1]]
-
-					# Group Combinations
-					grp.comb <- paste(combs[1,x], ':', combs[2,x])
-
-					# Collect all statistics into list
-					stats <- list(grp.comb, mean.diff, se, t, df, p, upper.conf, lower.conf)
-				  })
-
-				  # Unlist statistics collected earlier
-				  stats.unlisted <- lapply(statistics, function(x) {
-					unlist(x)
-				  })
-
-				  # Create dataframe from flattened list
-				  results <- data.frame(matrix(unlist(stats.unlisted), nrow = length(stats.unlisted), byrow=TRUE))
-
-				  # Select columns set as factors that should be numeric and change with as.numeric
-				  results[c(2, 3:ncol(results))] <- round(as.numeric(as.matrix(results[c(2, 3:ncol(results))])), digits = 3)
-
-				  # Rename data frame columns
-				  colnames(results) <- c('groups', 'Mean Difference', 'Standard Error', 't', 'df', 'p', 'upper limit', 'lower limit')
-
-				  return(results)
-				}
-
-				''')
-
-
-		self.gh = robjects.globalenv['games.howell']
 
 
 	def gamesHowellBinomial(self, groups):
@@ -92,34 +16,8 @@ def gamesHowellBinomial(self, groups):
 			groupNames+=int(groups[group][True]+groups[group][False])*[group]
 			groupValues+=int(groups[group][True])*[1]+int(groups[group][False])*[0]
 
-		results=self.gh(robjects.StrVector(groupNames),robjects.IntVector(groupValues) )
-		tableDict=stats.rTableToDict(results)
-		for group in groups:
-			tableDict[group]['rate']=groups[group][True]/(groups[group][False]+groups[group][True])
-		return tableDict
-
-
-
-	def rTableToDict(results):
-		"""Convert r results into a python dictionary format
-		Args:
-			results(obj): r object results in table format
-		Example:
-			stats.rTableToDict(self.gh(robjects.StrVector(groupNames),robjects.IntVector(groupValues)))
-		"""
-		tableDict={}
-		for i in range(len(results[0])):
-			firstQuant=str(results.rx(i+1,True)[0]).split("\n")[0].split(":")[0].split(' ')[1]
-			secondQuant=str(results.rx(i+1,True)[0]).split("\n")[0].split(": ")[1]
-			#print(orgOne)
-			#print(orgTwo)
-			pValue=str(results.rx(i+1,True)[5]).split(' ')[1].split("\n")[0];
-			if(firstQuant not in tableDict):
-				tableDict[firstQuant]={}
-			if(secondQuant not in tableDict):
-				tableDict[secondQuant]={}
+		df=pd.DataFrame()
+		df['group_names']=groupNames
+		df['group_values']=groupValues
 
-			tableDict[firstQuant][secondQuant]=pValue
-			tableDict[secondQuant][firstQuant]=pValue
-		#print(tableDict)
-		return tableDict;
+		return pairwise_gameshowell(dv='group_values', between='group_names', data=df)