tdt_result_t* tdt_result_new(char *chromosome, int chromosome_len, unsigned long int position, char *reference, int reference_len,
char *alternate, int alternate_len, double t1, double t2, double chi_square) {
tdt_result_t *result = (tdt_result_t*) malloc (sizeof(tdt_result_t));
result->chromosome = strndup(chromosome, chromosome_len);
result->position = position;
result->reference = strndup(reference, reference_len);
result->alternate = strndup(alternate, alternate_len);
result->t1 = t1;
result->t2 = t2;
result->odds_ratio = (t2 == 0.0) ? NAN : ((double) t1/t2);
result->chi_square = chi_square;
result->p_value = 1 - gsl_cdf_chisq_P(chi_square, 1);
return result;
}
assoc_basic_result_t* assoc_basic_result_new(char *chromosome, int chromosome_len, unsigned long int position,
char *reference, int reference_len, char *alternate, int alternate_len,
int affected1, int affected2, int unaffected1, int unaffected2, double chi_square) {
assoc_basic_result_t *result = (assoc_basic_result_t*) malloc (sizeof(assoc_basic_result_t));
result->chromosome = strndup(chromosome, chromosome_len);
result->position = position;
result->reference = strndup(reference, reference_len);
result->alternate = strndup(alternate, alternate_len);
result->affected1 = affected1;
result->affected2 = affected2;
result->unaffected1 = unaffected1;
result->unaffected2 = unaffected2;
result->odds_ratio = (affected2 == 0 || unaffected1 == 0) ? NAN :
((double) affected1 / affected2) * ((double) unaffected2 / unaffected1);
result->chi_square = chi_square;
result->p_value = 1 - gsl_cdf_chisq_P(chi_square, 1);
return result;
}
string GraphState::to_s ( unsigned int frequency ) {
float p, sig;
if (fm::chisq->active) {
if (!fm::regression) {
p = fm::chisq->p;
sig = fm::chisq->sig;
}
else {
p = fm::ks->p;
sig = fm::ks->sig;
}
}
if (!fm::chisq->active || p >= sig) {
string oss;
if (fm::gsp_out) {
to_s(oss);
return oss;
}
else {
if (fm::do_yaml) oss.append ("- [ ");
// output smarts
if (fm::do_yaml) oss.append ("\"");
int i;
for ( i = nodes.size()-1; i >= 0; i-- ) { // edges
if (nodes[i].edges.size()==1) break;
}
DfsOut(i, oss, i);
if (fm::do_yaml) oss.append ("\", ");
else oss.append("\t");
// output chisq
if (fm::chisq->active) {
if (fm::do_yaml) {
if (!fm::pvalues || fm::regression) {
char x[20];
sprintf(x,"%.4f", p);
(oss.append(x)).append(", ");
}
else {
char x[20];
sprintf(x,"%.4f", gsl_cdf_chisq_P(p, 1));
(oss.append(x)).append(", ");
}
}
}
// output freq
if (fm::chisq->active) {
if (!fm::regression && frequency != (fm::chisq->fa+fm::chisq->fi)) {
cerr << "Notice: Wrong counts for frequency " << frequency << " [!=" << fm::chisq->fa << "(fa)+" << fm::chisq->fi << "(fi)]." << endl;
}
}
else {
char x[20];
sprintf(x,"%i", frequency);
oss.append(x);
}
// output occurrences
if (fm::chisq->active) {
set<Tid> fa_set, fi_set;
if (!fm::regression) {
fa_set = fm::chisq->fa_set;
fi_set = fm::chisq->fi_set;
}
else {
fa_set = fm::ks->fa_set;
fi_set = fm::ks->fi_set;
}
oss.append ("[");
set<Tid>::iterator iter;
char x[20];
if (fm::do_yaml) {
set<Tid>::iterator begin = fa_set.begin();
set<Tid>::iterator end = fa_set.end();
set<Tid>::iterator last = end;
if (fa_set.size()) last = --(fa_set.end());
for (iter = begin; iter != end; iter++) {
if (iter != begin) oss.append (",");
oss.append (" ");
sprintf(x,"%i", (*iter));
oss.append (x);
if ((last != end) && (iter == last)) oss.append (" ");
}
if (!fm::regression) oss.append ("], [");
begin = fi_set.begin();
end = fi_set.end();
last = end;
if (fi_set.size()) last = --(fi_set.end());
for (iter = begin; iter != end; iter++) {
if (iter != begin) oss.append (",");
oss.append (" ");
sprintf(x,"%i", (*iter));
oss.append (x);
if ((last != end) && (iter == last)) oss.append (" ");
}
}
if (!fm::do_yaml) {
set<Tid> ids;
ids.insert(fa_set.begin(), fa_set.end());
ids.insert(fi_set.begin(), fi_set.end());
for (iter = ids.begin(); iter != ids.end(); iter++) {
sprintf(x,"%i", (*iter));
(oss.append (" ")).append(x);
}
}
if (!fm::do_yaml) oss.append (" ]");
else oss.append ("]");
}
if (fm::do_yaml) oss.append (" ]");
fm::console_out ? oss.append ("\n") : oss.append ("");
return oss;
}
}
else return "";
}
void GraphState::print ( unsigned int frequency ) {
float p, sig;
if (fm::chisq->active) {
if (!fm::regression) {
p = fm::chisq->p;
sig = fm::chisq->sig;
}
else {
p = fm::ks->p;
sig = fm::ks->sig;
}
}
if (!fm::chisq->active || p >= sig) {
if (fm::gsp_out) {
print(stdout);
}
else {
if (fm::do_yaml) {
putchar('-');
putchar(' ');
putchar('[');
putchar(' ');
putchar('"');
}
// output smarts
int i;
for ( i = nodes.size()-1; i >= 0; i-- ) { // edges
if (nodes[i].edges.size()==1) break;
}
DfsOut(i, i);
if (fm::do_yaml) {
putchar('"');
putchar(',');
putchar(' ');
}
// output chisq
if (fm::chisq->active) {
if (fm::do_yaml) {
if (!fm::pvalues || fm::regression) printf("%.4f, ", p);
else printf("%.4f, ", gsl_cdf_chisq_P(p,1));
}
else putchar('\t');
}
// output freq
if (fm::chisq->active) {
if (!fm::regression && frequency != (fm::chisq->fa+fm::chisq->fi)) {
cerr << "Error: wrong counts! " << frequency << "!=" << fm::chisq->fa + fm::chisq->fi << "(" << fm::chisq->fa << "+" << fm::chisq->fi << ")" << endl;
}
}
else {
if (fm::do_yaml) {
printf("%i", frequency);
}
else {
printf("%i\t", frequency);
};
}
// output occurrences
if (fm::chisq->active) {
set<Tid> fa_set, fi_set;
if (!fm::regression) {
fa_set = fm::chisq->fa_set;
fi_set = fm::chisq->fi_set;
}
else {
fa_set = fm::ks->fa_set;
fi_set = fm::ks->fi_set;
}
putchar ('[');
set<Tid>::iterator iter;
if (fm::do_yaml) {
for (iter = fa_set.begin(); iter != fa_set.end(); iter++) {
if (iter != fa_set.begin()) putchar (',');
putchar (' ');
printf("%i", (*iter));
}
if (!fm::regression) {
putchar (']');
putchar (',');
putchar (' ');
putchar ('[');
}
}
if (fm::do_yaml) {
for (iter = fi_set.begin(); iter != fi_set.end(); iter++) {
if (iter != fi_set.begin()) putchar (',');
printf(" %i", (*iter));
}
}
if (!fm::do_yaml) {
set<Tid> ids;
ids.insert(fa_set.begin(), fa_set.end());
ids.insert(fi_set.begin(), fi_set.end());
for (iter = ids.begin(); iter != ids.end(); iter++) {
putchar(' ');
printf("%i", (*iter));
}
}
putchar(' ');
putchar(']');
}
if (fm::do_yaml) {
putchar(' ');
putchar(']');
}
if(fm::console_out) putchar('\n');
}
}
}
void postprocessing()
{
long int i,num;
int j,k,l,s,isNew;
double eva,p_value;
int* loci;
loci=new int[iEpiModel];
vector<double> record;
record.resize(iEpiModel+2);
cout<<"Post-processing"<<endl;
cout<<"Dealing with top ranking SNP sets"<<endl;
num=comb_num(iLociModel,iEpiModel);
cout<<"Number of interatcions evaluated: "<<num*iTopModel<<endl;
int** combination;
combination=new int*[num];
for(i=0;i<num;i++)
combination[i]=new int[iEpiModel];
combination=throughout(iLociModel,iEpiModel,combination);
for(i=iTopModel-1;i>=0;i--)
{
for(j=0;j<num;j++)
{
for(k=0;k<iEpiModel;k++)
{
loci[k]=loci_TopModel[i][combination[j][k]];
}
eva=chi_square(loci,iEpiModel);
p_value=1-gsl_cdf_chisq_P(eva,pow(3,iEpiModel)-1);
if(p_value<pvalue)
{
isNew=1;
for(l=0;l<interactions.size();l++)
{
if(fabs(eva-interactions[l][iEpiModel])<0.00000001)
{
isNew=0;
break;
}
}
if(isNew)
{
for(s=0;s<iEpiModel;s++)
{
record[s]=(double)loci[s];
}
record[iEpiModel]=eva;
record[iEpiModel+1]=p_value;
interactions.push_back(record);
}
}
}
}
for(i=0;i<num;i++)
delete []combination[i];
delete []combination;
cout<<"Dealing with top ranking loci"<<endl;
num=comb_num(iTopLoci,iEpiModel);
//cout<<"hello"<<endl;
int** combination1;
combination1=new int*[num];
for(i=0;i<num;i++)
combination1[i]=new int[iEpiModel];
combination1=throughout(iTopLoci,iEpiModel,combination1);
cout<<"Number of interactions evaluated: "<<num<<endl;
for(i=0;i<num;i++)
{
for(j=0;j<iEpiModel;j++)
{
loci[j]=loci_TopLoci[combination1[i][j]];
//cout<<loci[j]<<" ";
}
//cout<<endl;
eva=chi_square(loci,iEpiModel);
p_value=1-gsl_cdf_chisq_P(eva,pow(3,iEpiModel)-1);
if(p_value<pvalue)
{
isNew=1;
for(l=0;l<interactions.size();l++)
{
if(fabs(eva-interactions[l][iEpiModel])<0.00000001)
{
isNew=0;
break;
}
}
if(isNew)
{
for(s=0;s<iEpiModel;s++)
{
record[s]=(double)loci[s];
}
record[iEpiModel]=eva;
record[iEpiModel+1]=p_value;
interactions.push_back(record);
}
}
}
for(i=0;i<num;i++)
delete []combination1[i];
delete []combination1;
}
double pp_model::calculate_waiting_times_log_predictive_df( double increment, bool lower_tail, bool two_sided, bool increment_parameters ){
double sum_log_pvals = 0;// sum_log_pvals2 = 0;
unsigned long long int how_many = 0;
unsigned long long int i2 = m_data_cont ? m_data_cont->find_data_index(m_current_t+increment,0,m_current_data_index) : (unsigned long long int)(m_current_t+increment);
double current_t = m_current_t;
if(i2>m_current_data_index){
while(i2>m_current_data_index){
double t = m_pp_time_scale ? m_pp_time_scale->cumulative_function( current_t, (*m_data_cont)[0][m_current_data_index] ) : (*m_data_cont)[0][m_current_data_index] - current_t;
m_log_predictive_df = calculate_log_posterior_predictive_pdf(t,0);//upper tail
if(t<=0){
cerr << "Rounding errors in event times"<< endl;
exit(1);
}
if(lower_tail||two_sided){
m_log_predictive_df2 = log(1-exp(m_log_predictive_df));//lower tail
if(lower_tail)
m_log_predictive_df = m_log_predictive_df2;
}
/* if(two_sided){//two-sided p-values version two - combine at the start
if(m_log_predictive_df > m_log_predictive_df2)
m_log_predictive_df = m_log_predictive_df2;
m_log_predictive_df += LOG_TWO;
}*/
sum_log_pvals += m_log_predictive_df;
// if(two_sided)//two-sided p-values version one - combine at the end
// sum_log_pvals2 += m_log_predictive_df2;
m_r++;
m_t += t;
m_alpha_star++;
m_beta_star += t;
how_many++;
current_t = (*m_data_cont)[0][m_current_data_index];
m_current_data_index++;
}
}
double t = (m_pp_time_scale&&m_data_cont) ? m_pp_time_scale->cumulative_function( current_t, m_current_t+increment ) : m_current_t+increment - current_t;
if(t>0){
m_log_predictive_df = calculate_log_posterior_predictive_pdf(t,0) - LOG_TWO;//upper tail
if(lower_tail||two_sided){
m_log_predictive_df2 = log(1-exp(m_log_predictive_df));//lower tail
if(lower_tail)
m_log_predictive_df = m_log_predictive_df2;
}
/* if(two_sided){
if(m_log_predictive_df > m_log_predictive_df2)
m_log_predictive_df = m_log_predictive_df2;
m_log_predictive_df += LOG_TWO;
}*/
sum_log_pvals += m_log_predictive_df;
// if(two_sided)
// sum_log_pvals2 += m_log_predictive_df2;
m_t += t;
m_beta_star += t;
how_many++;
}
// if(two_sided && sum_log_pvals2 < sum_log_pvals)
// sum_log_pvals = sum_log_pvals2;
m_currently_observable = how_many > 0;
if(m_currently_observable){
m_log_predictive_df = how_many==1 ? sum_log_pvals : log(gsl_cdf_chisq_Q(-2*sum_log_pvals, 2*how_many));
// if(two_sided)
// m_log_predictive_df += LOG_TWO;
if(two_sided){
if(m_log_predictive_df>=0)//rounding erros
m_log_predictive_df2 = log(gsl_cdf_chisq_P(-2*sum_log_pvals, 2*how_many));
else
m_log_predictive_df2 = log(1-exp(m_log_predictive_df));//lower tail
if(m_log_predictive_df > m_log_predictive_df2)
m_log_predictive_df = m_log_predictive_df2;
m_log_predictive_df += LOG_TWO;
}
}
else
m_log_predictive_df = -LOG_TWO;
m_pvalue_pair = make_pair(m_log_predictive_df,m_log_predictive_df);
m_pvalue_pair_on_log_scale = true;
return m_log_predictive_df;
}
