Assistance with Analyzing Repeated Measures Data in Udder Health Study

[Aminu Shittu]

Dear StataListers,

I have collected milk samples from the individual quarters of 127 lactating cows across 14 farms to assess udder health (mammary gland condition). In total, 508 samples (127 cows x 4 quarters) were collected from farms with varying number of lactating cows and other farm-level characteristics. My response variable is dichotomous, indicating the presence or absence of subclinical mastitis (SCM).

Each cow has 4 rows in the dataset corresponding to the quarters (LFQ, LHQ, RFQ, RHQ), all linked to their respective cow_ and farm_ids. Additionally, cow-level variables (e.g., age (similar within a cow), breed (similar within a cow), BCS (similar within a cow), teat characteristics (varies within a cow), etc) were also recorded for each quarter, resulting in 4 rows per cow ion the dataset.

Since I am encountering this kind of dataset with repeated measures (four times per cow), I am seeking guidance on analyzing udder and cow characteristics as predictors of SCM. I hope the information I've provided gives you a clear idea of my dataset structure, a subset is pasted below. Any references or links to relevant resources for this type of analysis (from the descriptive analysis to advance statistical modelling) would also be greatly appreciated. I am using Stata IC v.15.

| cow_id | farm_id | quarter | SCM | age | breed | BCS |
|--------|---------|---------|-----|-----|--------|-----|
| Cow1 | Farm1 | LFQ | 1 | 3.0 | Breed A| 2.5 |
| Cow1 | Farm1 | LHQ | 0 | 3.0 | Breed A| 2.5 |
| Cow1 | Farm1 | RFQ | 0 | 3.0 | Breed A| 2.5 |
| Cow1 | Farm1 | RHQ | 1 | 3.0 | Breed A| 2.5 |
| Cow2 | Farm1 | LFQ | 0 | 4.0 | Breed B| 3.0 |
| Cow2 | Farm1 | LHQ | 1 | 4.0 | Breed B| 3.0 |
| Cow2 | Farm1 | RFQ | 1 | 4.0 | Breed B| 3.0 |
| Cow2 | Farm1 | RHQ | 0 | 4.0 | Breed B| 3.0 |
| Cow3 | Farm2 | LFQ | 1 | 5.0 | Breed A| 2.8 |
| Cow3 | Farm2 | LHQ | 1 | 5.0 | Breed A| 2.8 |
| Cow3 | Farm2 | RFQ | 0 | 5.0 | Breed A| 2.8 |
| Cow3 | Farm2 | RHQ | 0 | 5.0 | Breed A| 2.8 |

Thank you in advance for your assistance.

Best regards,

Aminu Shittu

Faculty of Veterinary Medicine,
Usmanu Danfodiyo University Sokoto,
P.M.B 2254, Sokoto State,
NIGERIA.

[Erik Ruzek]

Aminu,

Given your interest in modeling variables that vary within cows and those that are constant (similar) within a cow, you are going to want to look at mixed effects models for binary outcomes (melogit). Stata's PDF documentation on these models is a very good place to start. If you are not familiar with analyzing repeated measures data using mixed models, then reading the introduction section of the me PDF documentation is also suggested.

[Aminu Shittu]

Thank you so much, Erik, and I apologize for the delayed response. I will review the document you've shared. Best regards, Aminu.