Hello. I work at a high school and while I took a stats class many years ago using SPSS, its been a long time and I have forgotten a lot. I have a dataset in which we want to look at the difference between reaction time between students' fastest and slowest reading times. First, I want to know if these times are significantly different from one another. How do I do this? Then I have some co-variates like age and comprehension I'd like to include. Do I do this as a univariate model or linear regression? If there is not a significant difference, should I even proceed with further analysis? I'd like to know more how the fastest times may have been impacted by the co-variates.

Hi. You could use COMPUTE to create a "difference score" for each student (fastest - slowest, for example):

COMPUTE diff=fastest-slowest.

and then make that your dependent variable in a linear regression, using your covariates as the independent variables.

Alternatively (and equivalently in this case), you could set up a repeated measures analysis of variance (in GLM) with a within-subjects variable (call it "time", say) along with the covariates in the model. 

Either way, you will get the same basic answer.

------------------------------
Rick Marcantonio
Quality Assurance
IBM

Original Message:
Sent: Wed February 15, 2023 02:05 PM
From: Stacy Adams
Subject: Analysis Suggestions for a Stats Novice

Hello. I work at a high school and while I took a stats class many years ago using SPSS, its been a long time and I have forgotten a lot. I have a dataset in which we want to look at the difference between reaction time between students' fastest and slowest reading times. First, I want to know if these times are significantly different from one another. How do I do this? Then I have some co-variates like age and comprehension I'd like to include. Do I do this as a univariate model or linear regression? If there is not a significant difference, should I even proceed with further analysis? I'd like to know more how the fastest times may have been impacted by the co-variates.

------------------------------
Stacy Adams
------------------------------

Thank you, Rick. Does that also help me answer whether the fastest and slowest times are significantly different from one another? If I do that and add average speed in a linear regression model and it significantly predicts the diff variable, how do I interpret that? I did a Chi-Square test of the two means and its not significant, but I don't know if that is right.

I also just tried both linear regression and a univariate model with fastest speed as the dependent and slowest as an independent. The linear regression model is significant but not in the univariate regression model. 

I had also tried adding fastest in a univariate model and as soon as I start adding covariates like average speed, age, comprehension, I get a bunch of . instead of results. Am I just doing this all wrong?

------------------------------
Stacy Adams

Original Message:
Sent: Wed February 15, 2023 03:42 PM
From: Rick Marcantonio
Subject: Analysis Suggestions for a Stats Novice

Hi. You could use COMPUTE to create a "difference score" for each student (fastest - slowest, for example):

COMPUTE diff=fastest-slowest.

and then make that your dependent variable in a linear regression, using your covariates as the independent variables.

Alternatively (and equivalently in this case), you could set up a repeated measures analysis of variance (in GLM) with a within-subjects variable (call it "time", say) along with the covariates in the model. 

Either way, you will get the same basic answer.

------------------------------
Rick Marcantonio
Quality Assurance
IBM

Original Message:
Sent: Wed February 15, 2023 02:05 PM
From: Stacy Adams
Subject: Analysis Suggestions for a Stats Novice

Hello. I work at a high school and while I took a stats class many years ago using SPSS, its been a long time and I have forgotten a lot. I have a dataset in which we want to look at the difference between reaction time between students' fastest and slowest reading times. First, I want to know if these times are significantly different from one another. How do I do this? Then I have some co-variates like age and comprehension I'd like to include. Do I do this as a univariate model or linear regression? If there is not a significant difference, should I even proceed with further analysis? I'd like to know more how the fastest times may have been impacted by the co-variates.

------------------------------
Stacy Adams
------------------------------

When I said the models were equivalent, what I meant was a linear regression with time difference as the dependent measure compared to a repeated-measures with two dependent variables (one within-subjects variable, "time"): fastest and slowest speeds.

The univariate model with fastest time only is answering the different question: "Is the fastest speed significantly different from 0?" My guess is that it would be.

If you want to know whether the average fastest speed is significantly different from the average slowest speed, then do a paired t-test of slowest and fastest, which tests the null hypothesis that the difference is 0. Equivalently, you could do a single-sample t-test of the time difference variable I talked about against a null value of 0.

When you start entering covariates, then you need the linear regression (or repeated-measures).

About whether you're doing it wrong, I'd need to see your data and syntax. 

Send it to marcantr@us.ibm.com if you want, and I'll look at it.

------------------------------
Rick Marcantonio
Quality Assurance
IBM

Original Message:
Sent: Wed February 15, 2023 04:18 PM
From: Stacy Adams
Subject: Analysis Suggestions for a Stats Novice

Thank you, Rick. Does that also help me answer whether the fastest and slowest times are significantly different from one another? If I do that and add average speed in a linear regression model and it significantly predicts the diff variable, how do I interpret that? I did a Chi-Square test of the two means and its not significant, but I don't know if that is right.

I also just tried both linear regression and a univariate model with fastest speed as the dependent and slowest as an independent. The linear regression model is significant but not in the univariate regression model. 

I had also tried adding fastest in a univariate model and as soon as I start adding covariates like average speed, age, comprehension, I get a bunch of . instead of results. Am I just doing this all wrong?

------------------------------
Stacy Adams

Original Message:
Sent: Wed February 15, 2023 03:42 PM
From: Rick Marcantonio
Subject: Analysis Suggestions for a Stats Novice

Hi. You could use COMPUTE to create a "difference score" for each student (fastest - slowest, for example):

COMPUTE diff=fastest-slowest.

and then make that your dependent variable in a linear regression, using your covariates as the independent variables.

Alternatively (and equivalently in this case), you could set up a repeated measures analysis of variance (in GLM) with a within-subjects variable (call it "time", say) along with the covariates in the model. 

Either way, you will get the same basic answer.

------------------------------
Rick Marcantonio
Quality Assurance
IBM

Original Message:
Sent: Wed February 15, 2023 02:05 PM
From: Stacy Adams
Subject: Analysis Suggestions for a Stats Novice

Hello. I work at a high school and while I took a stats class many years ago using SPSS, its been a long time and I have forgotten a lot. I have a dataset in which we want to look at the difference between reaction time between students' fastest and slowest reading times. First, I want to know if these times are significantly different from one another. How do I do this? Then I have some co-variates like age and comprehension I'd like to include. Do I do this as a univariate model or linear regression? If there is not a significant difference, should I even proceed with further analysis? I'd like to know more how the fastest times may have been impacted by the co-variates.

------------------------------
Stacy Adams
------------------------------

Okay, so I can just enter fastest and slowest as 1 pair and run? In this case, it is significant. An additional question, first I had run a one-way ANOVA to compare which was not significant. Is ANOVA for comparing groups?

When I said the models were equivalent, what I meant was a linear regression with time difference as the dependent measure compared to a repeated-measures with two dependent variables (one within-subjects variable, "time"): fastest and slowest speeds.

The univariate model with fastest time only is answering the different question: "Is the fastest speed significantly different from 0?" My guess is that it would be.

If you want to know whether the average fastest speed is significantly different from the average slowest speed, then do a paired t-test of slowest and fastest, which tests the null hypothesis that the difference is 0. Equivalently, you could do a single-sample t-test of the time difference variable I talked about against a null value of 0.

When you start entering covariates, then you need the linear regression (or repeated-measures).

About whether you're doing it wrong, I'd need to see your data and syntax. 

Send it to marcantr@us.ibm.com if you want, and I'll look at it.

------------------------------
Rick Marcantonio
Quality Assurance
IBM

Original Message:
Sent: Wed February 15, 2023 04:18 PM
From: Stacy Adams
Subject: Analysis Suggestions for a Stats Novice

Thank you, Rick. Does that also help me answer whether the fastest and slowest times are significantly different from one another? If I do that and add average speed in a linear regression model and it significantly predicts the diff variable, how do I interpret that? I did a Chi-Square test of the two means and its not significant, but I don't know if that is right.

I also just tried both linear regression and a univariate model with fastest speed as the dependent and slowest as an independent. The linear regression model is significant but not in the univariate regression model. 

I had also tried adding fastest in a univariate model and as soon as I start adding covariates like average speed, age, comprehension, I get a bunch of . instead of results. Am I just doing this all wrong?

------------------------------
Stacy Adams

Original Message:
Sent: Wed February 15, 2023 03:42 PM
From: Rick Marcantonio
Subject: Analysis Suggestions for a Stats Novice

Hi. You could use COMPUTE to create a "difference score" for each student (fastest - slowest, for example):

COMPUTE diff=fastest-slowest.

and then make that your dependent variable in a linear regression, using your covariates as the independent variables.

Alternatively (and equivalently in this case), you could set up a repeated measures analysis of variance (in GLM) with a within-subjects variable (call it "time", say) along with the covariates in the model. 

Either way, you will get the same basic answer.

------------------------------
Rick Marcantonio
Quality Assurance
IBM

Original Message:
Sent: Wed February 15, 2023 02:05 PM
From: Stacy Adams
Subject: Analysis Suggestions for a Stats Novice

Hello. I work at a high school and while I took a stats class many years ago using SPSS, its been a long time and I have forgotten a lot. I have a dataset in which we want to look at the difference between reaction time between students' fastest and slowest reading times. First, I want to know if these times are significantly different from one another. How do I do this? Then I have some co-variates like age and comprehension I'd like to include. Do I do this as a univariate model or linear regression? If there is not a significant difference, should I even proceed with further analysis? I'd like to know more how the fastest times may have been impacted by the co-variates.

------------------------------
Stacy Adams
------------------------------

Let me be sure that I'm clear:

"Okay, so I can just enter fastest and slowest as 1 pair and run? In this case, it is significant."

Just to be sure, that test looks something like this:

T-TEST PAIRS=slowest WITH fastest (PAIRED).

right?

"...first I had run a one-way ANOVA to compare which was not significant. Is ANOVA for comparing groups?"

The analysis of variance is a general model type whose purpose is to compare sample means. There can be groups - there often are - but there can be one group measured at 2 or more times, or some combination of both groups and multiple times.

When I hear "one-way ANOVA," I think of two or more independent groups; for example, students in class 1 compared to other students in class 2. From your description, I didn't see any independent groups. I did see one group - students - who gave two scores. That makes it an ANOVA but not a one-way as I think of it, but as a repeated measures - the simplest of which is T-TEST PAIRS. If you add covariates to it, it becomes a GLM repeated-measures.

Okay, so I can just enter fastest and slowest as 1 pair and run? In this case, it is significant. An additional question, first I had run a one-way ANOVA to compare which was not significant. Is ANOVA for comparing groups?

When I said the models were equivalent, what I meant was a linear regression with time difference as the dependent measure compared to a repeated-measures with two dependent variables (one within-subjects variable, "time"): fastest and slowest speeds.

The univariate model with fastest time only is answering the different question: "Is the fastest speed significantly different from 0?" My guess is that it would be.

If you want to know whether the average fastest speed is significantly different from the average slowest speed, then do a paired t-test of slowest and fastest, which tests the null hypothesis that the difference is 0. Equivalently, you could do a single-sample t-test of the time difference variable I talked about against a null value of 0.

When you start entering covariates, then you need the linear regression (or repeated-measures).

About whether you're doing it wrong, I'd need to see your data and syntax. 

Send it to marcantr@us.ibm.com if you want, and I'll look at it.

------------------------------
Rick Marcantonio
Quality Assurance
IBM

Original Message:
Sent: Wed February 15, 2023 04:18 PM
From: Stacy Adams
Subject: Analysis Suggestions for a Stats Novice

Thank you, Rick. Does that also help me answer whether the fastest and slowest times are significantly different from one another? If I do that and add average speed in a linear regression model and it significantly predicts the diff variable, how do I interpret that? I did a Chi-Square test of the two means and its not significant, but I don't know if that is right.

I also just tried both linear regression and a univariate model with fastest speed as the dependent and slowest as an independent. The linear regression model is significant but not in the univariate regression model. 

I had also tried adding fastest in a univariate model and as soon as I start adding covariates like average speed, age, comprehension, I get a bunch of . instead of results. Am I just doing this all wrong?

------------------------------
Stacy Adams

Original Message:
Sent: Wed February 15, 2023 03:42 PM
From: Rick Marcantonio
Subject: Analysis Suggestions for a Stats Novice

Hi. You could use COMPUTE to create a "difference score" for each student (fastest - slowest, for example):

COMPUTE diff=fastest-slowest.

and then make that your dependent variable in a linear regression, using your covariates as the independent variables.

Alternatively (and equivalently in this case), you could set up a repeated measures analysis of variance (in GLM) with a within-subjects variable (call it "time", say) along with the covariates in the model. 

Either way, you will get the same basic answer.

------------------------------
Rick Marcantonio
Quality Assurance
IBM

Original Message:
Sent: Wed February 15, 2023 02:05 PM
From: Stacy Adams
Subject: Analysis Suggestions for a Stats Novice

Hello. I work at a high school and while I took a stats class many years ago using SPSS, its been a long time and I have forgotten a lot. I have a dataset in which we want to look at the difference between reaction time between students' fastest and slowest reading times. First, I want to know if these times are significantly different from one another. How do I do this? Then I have some co-variates like age and comprehension I'd like to include. Do I do this as a univariate model or linear regression? If there is not a significant difference, should I even proceed with further analysis? I'd like to know more how the fastest times may have been impacted by the co-variates.

------------------------------
Stacy Adams
------------------------------

When I said the models were equivalent, what I meant was a linear regression with time difference as the dependent measure compared to a repeated-measures with two dependent variables (one within-subjects variable, "time"): fastest and slowest speeds.

The univariate model with fastest time only is answering the different question: "Is the fastest speed significantly different from 0?" My guess is that it would be.

If you want to know whether the average fastest speed is significantly different from the average slowest speed, then do a paired t-test of slowest and fastest, which tests the null hypothesis that the difference is 0. Equivalently, you could do a single-sample t-test of the time difference variable I talked about against a null value of 0.

When you start entering covariates, then you need the linear regression (or repeated-measures).

About whether you're doing it wrong, I'd need to see your data and syntax. 

Send it to marcantr@us.ibm.com if you want, and I'll look at it.

------------------------------
Rick Marcantonio
Quality Assurance
IBM

Original Message:
Sent: Wed February 15, 2023 04:18 PM
From: Stacy Adams
Subject: Analysis Suggestions for a Stats Novice

Thank you, Rick. Does that also help me answer whether the fastest and slowest times are significantly different from one another? If I do that and add average speed in a linear regression model and it significantly predicts the diff variable, how do I interpret that? I did a Chi-Square test of the two means and its not significant, but I don't know if that is right.

I also just tried both linear regression and a univariate model with fastest speed as the dependent and slowest as an independent. The linear regression model is significant but not in the univariate regression model. 

I had also tried adding fastest in a univariate model and as soon as I start adding covariates like average speed, age, comprehension, I get a bunch of . instead of results. Am I just doing this all wrong?

------------------------------
Stacy Adams

Original Message:
Sent: Wed February 15, 2023 03:42 PM
From: Rick Marcantonio
Subject: Analysis Suggestions for a Stats Novice

Hi. You could use COMPUTE to create a "difference score" for each student (fastest - slowest, for example):

COMPUTE diff=fastest-slowest.

and then make that your dependent variable in a linear regression, using your covariates as the independent variables.

Alternatively (and equivalently in this case), you could set up a repeated measures analysis of variance (in GLM) with a within-subjects variable (call it "time", say) along with the covariates in the model. 

Either way, you will get the same basic answer.

------------------------------
Rick Marcantonio
Quality Assurance
IBM

Original Message:
Sent: Wed February 15, 2023 02:05 PM
From: Stacy Adams
Subject: Analysis Suggestions for a Stats Novice

Hello. I work at a high school and while I took a stats class many years ago using SPSS, its been a long time and I have forgotten a lot. I have a dataset in which we want to look at the difference between reaction time between students' fastest and slowest reading times. First, I want to know if these times are significantly different from one another. How do I do this? Then I have some co-variates like age and comprehension I'd like to include. Do I do this as a univariate model or linear regression? If there is not a significant difference, should I even proceed with further analysis? I'd like to know more how the fastest times may have been impacted by the co-variates.

------------------------------
Stacy Adams
------------------------------

Hi @Rick Marcantonio

 I have a similar question, except, I'm trying to determine whether 5 IVs are significantly predicting the number of projects completed over 12 months for 21 teams. 7 of the 12 months had 0 projects completed for several teams. My questions are: 

1. I tried using ZEROINFL, but one of the months received the output Warning, "NaNs"? How do I fix this?
2. I can't quite figure out if I'm using the right analyses to predict the outcome variable. I don't think I need to run a multilevel GLM because I already aggregated all the variables to the team level. Also, there is no data available for each team member's contribution to the completion of each project. However, I do have individual response data for all the IVs. Do I still need to conduct a multilevel regression? If not, why? If so, then how? Do I need to use a "Growth Model"? 

Thank you!!

When I said the models were equivalent, what I meant was a linear regression with time difference as the dependent measure compared to a repeated-measures with two dependent variables (one within-subjects variable, "time"): fastest and slowest speeds.

The univariate model with fastest time only is answering the different question: "Is the fastest speed significantly different from 0?" My guess is that it would be.

If you want to know whether the average fastest speed is significantly different from the average slowest speed, then do a paired t-test of slowest and fastest, which tests the null hypothesis that the difference is 0. Equivalently, you could do a single-sample t-test of the time difference variable I talked about against a null value of 0.

When you start entering covariates, then you need the linear regression (or repeated-measures).

About whether you're doing it wrong, I'd need to see your data and syntax. 

Send it to marcantr@us.ibm.com if you want, and I'll look at it.

------------------------------
Rick Marcantonio
Quality Assurance
IBM

Original Message:
Sent: Wed February 15, 2023 04:18 PM
From: Stacy Adams
Subject: Analysis Suggestions for a Stats Novice

Thank you, Rick. Does that also help me answer whether the fastest and slowest times are significantly different from one another? If I do that and add average speed in a linear regression model and it significantly predicts the diff variable, how do I interpret that? I did a Chi-Square test of the two means and its not significant, but I don't know if that is right.

I also just tried both linear regression and a univariate model with fastest speed as the dependent and slowest as an independent. The linear regression model is significant but not in the univariate regression model. 

I had also tried adding fastest in a univariate model and as soon as I start adding covariates like average speed, age, comprehension, I get a bunch of . instead of results. Am I just doing this all wrong?

------------------------------
Stacy Adams

Original Message:
Sent: Wed February 15, 2023 03:42 PM
From: Rick Marcantonio
Subject: Analysis Suggestions for a Stats Novice

Hi. You could use COMPUTE to create a "difference score" for each student (fastest - slowest, for example):

COMPUTE diff=fastest-slowest.

and then make that your dependent variable in a linear regression, using your covariates as the independent variables.

Alternatively (and equivalently in this case), you could set up a repeated measures analysis of variance (in GLM) with a within-subjects variable (call it "time", say) along with the covariates in the model. 

Either way, you will get the same basic answer.

------------------------------
Rick Marcantonio
Quality Assurance
IBM

Original Message:
Sent: Wed February 15, 2023 02:05 PM
From: Stacy Adams
Subject: Analysis Suggestions for a Stats Novice

Hello. I work at a high school and while I took a stats class many years ago using SPSS, its been a long time and I have forgotten a lot. I have a dataset in which we want to look at the difference between reaction time between students' fastest and slowest reading times. First, I want to know if these times are significantly different from one another. How do I do this? Then I have some co-variates like age and comprehension I'd like to include. Do I do this as a univariate model or linear regression? If there is not a significant difference, should I even proceed with further analysis? I'd like to know more how the fastest times may have been impacted by the co-variates.

------------------------------
Stacy Adams
------------------------------

Hi @Rick Marcantonio

 I have a similar question, except, I'm trying to determine whether 5 IVs are significantly predicting the number of projects completed over 12 months for 21 teams. 7 of the 12 months had 0 projects completed for several teams. My questions are: 

1. I tried using ZEROINFL, but one of the months received the output Warning, "NaNs"? How do I fix this?
2. I can't quite figure out if I'm using the right analyses to predict the outcome variable. I don't think I need to run a multilevel GLM because I already aggregated all the variables to the team level. Also, there is no data available for each team member's contribution to the completion of each project. However, I do have individual response data for all the IVs. Do I still need to conduct a multilevel regression? If not, why? If so, then how? Do I need to use a "Growth Model"? 

Thank you!!

When I said the models were equivalent, what I meant was a linear regression with time difference as the dependent measure compared to a repeated-measures with two dependent variables (one within-subjects variable, "time"): fastest and slowest speeds.

The univariate model with fastest time only is answering the different question: "Is the fastest speed significantly different from 0?" My guess is that it would be.

If you want to know whether the average fastest speed is significantly different from the average slowest speed, then do a paired t-test of slowest and fastest, which tests the null hypothesis that the difference is 0. Equivalently, you could do a single-sample t-test of the time difference variable I talked about against a null value of 0.

When you start entering covariates, then you need the linear regression (or repeated-measures).

About whether you're doing it wrong, I'd need to see your data and syntax. 

Send it to marcantr@us.ibm.com if you want, and I'll look at it.

------------------------------
Rick Marcantonio
Quality Assurance
IBM

Original Message:
Sent: Wed February 15, 2023 04:18 PM
From: Stacy Adams
Subject: Analysis Suggestions for a Stats Novice

Thank you, Rick. Does that also help me answer whether the fastest and slowest times are significantly different from one another? If I do that and add average speed in a linear regression model and it significantly predicts the diff variable, how do I interpret that? I did a Chi-Square test of the two means and its not significant, but I don't know if that is right.

I also just tried both linear regression and a univariate model with fastest speed as the dependent and slowest as an independent. The linear regression model is significant but not in the univariate regression model. 

I had also tried adding fastest in a univariate model and as soon as I start adding covariates like average speed, age, comprehension, I get a bunch of . instead of results. Am I just doing this all wrong?

------------------------------
Stacy Adams

Original Message:
Sent: Wed February 15, 2023 03:42 PM
From: Rick Marcantonio
Subject: Analysis Suggestions for a Stats Novice

Hi. You could use COMPUTE to create a "difference score" for each student (fastest - slowest, for example):

COMPUTE diff=fastest-slowest.

and then make that your dependent variable in a linear regression, using your covariates as the independent variables.

Alternatively (and equivalently in this case), you could set up a repeated measures analysis of variance (in GLM) with a within-subjects variable (call it "time", say) along with the covariates in the model. 

Either way, you will get the same basic answer.

------------------------------
Rick Marcantonio
Quality Assurance
IBM

Original Message:
Sent: Wed February 15, 2023 02:05 PM
From: Stacy Adams
Subject: Analysis Suggestions for a Stats Novice

Hello. I work at a high school and while I took a stats class many years ago using SPSS, its been a long time and I have forgotten a lot. I have a dataset in which we want to look at the difference between reaction time between students' fastest and slowest reading times. First, I want to know if these times are significantly different from one another. How do I do this? Then I have some co-variates like age and comprehension I'd like to include. Do I do this as a univariate model or linear regression? If there is not a significant difference, should I even proceed with further analysis? I'd like to know more how the fastest times may have been impacted by the co-variates.

------------------------------
Stacy Adams
------------------------------